MCP tool reference

Generated from the tool registry (packages/agent). 108 tools. Connect with the same API token you use for the REST API; scopes gate each tool exactly as they gate the REST route it maps to. Every mutating call is audited (Settings → Agents → Activity, or list_agent_activity).

Account

whoami

Who am I? Returns how this connection is authenticated (API token or session), the organization(s) it can reach, the granted scopes, and the user behind it. Call this first. Works with any authenticated connection regardless of scopes.

• Scope: org:read · read-only
• REST: GET /v1/whoami

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "whoami",
    "arguments": {}
  }
}

Documentation

search_docs

Full-text search over Bugwatch's documentation (setup, SDKs, API, MCP, alerts, billing, security). Returns slugs to pass to get_doc.

• Scope: org:read · read-only
• Arguments:
  • query string (required) — Search terms
  • limit integer

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "search_docs",
    "arguments": {
      "query": "source maps"
    }
  }
}

get_doc

Read one documentation page as Markdown by slug (from search_docs or list_docs).

• Scope: org:read · read-only
• Arguments:
  • slug string (required) — Doc slug, e.g. quickstart

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_doc",
    "arguments": {
      "slug": "quickstart"
    }
  }
}

list_docs

Table of contents of the documentation: slug, title, one-line description per page.

• Scope: org:read · read-only

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_docs",
    "arguments": {}
  }
}

Projects

list_projects

List the projects in an organization with slug, name, platform, and dashboard link.

• Scope: project:read · read-only
• REST: GET /v1/orgs/{org}/projects
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_projects",
    "arguments": {
      "org": "acme"
    }
  }
}

get_project

Project settings: name, platform, retention tier, store-IP flag, inbound filters, project number.

• Scope: project:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_project",
    "arguments": {
      "org": "acme",
      "project": "checkout-api"
    }
  }
}

create_project

Create a project. Returns its DSN-bearing key; pass the platform id (e.g. javascript-nextjs, python-django) when known.

• Scope: project:write · mutating, audited
• REST: POST /v1/orgs/{org}/projects
• Arguments:
  • org string (required) — Organization slug
  • slug string (required) — lowercase letters, digits, dashes
  • name string (required)
  • platform string — Platform id from the registry (optional)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_project",
    "arguments": {
      "org": "acme",
      "slug": "checkout-api",
      "name": "Checkout API",
      "platform": "node"
    }
  }
}

update_project

Update project name, platform, store-IP flag, or inbound filters (crawlers, legacy browsers, localhost, extensions, message deny list, allowed domains).

• Scope: project:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/projects/{project}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • name string
  • platform string
  • storeIp boolean — Keep client IP on events (default false)
  • filters object — Inbound filters

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "update_project",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "filters": {
        "dropCrawlers": true
      }
    }
  }
}

get_repo_connection

Whether this organization has a repository connected for the fix agent, and which one. Connecting or disconnecting is human-only — it stores a credential that can write to source code.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/repo-connection
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_repo_connection",
    "arguments": {
      "org": "acme"
    }
  }
}

Issues

search_issues

List or search issues in a project (compact: id, level, status, title, impact, deep link). Filter by status, full-text q, sort, and page with cursor.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • q string — Free-text match on title/culprit
  • status string one of unresolved, resolved, ignored, regressed
  • sort string one of last_seen, first_seen
  • limit integer
  • cursor string — nextCursor from a previous call

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "search_issues",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "status": "unresolved",
      "limit": 10
    }
  }
}

get_issue

Compact summary of one issue (title, culprit, status, impact, first/last seen, deep link). For frames and source context call get_fix_context.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues/{shortId}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_issue",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

get_fix_context

Everything needed to FIX an issue in one payload (JSON, can be large): symbolicated in-app frames with source context, breadcrumbs, release + commit, impact numbers, similar resolved issues. Call before attempting a fix.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues/{shortId}/fix-context
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_fix_context",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

get_issue_history

Day-by-day events and estimated affected users for an issue, from the nightly rollup — so it reaches back FURTHER than the analytics window the issue chart uses (up to 400 days). Use it to answer whether an issue is new, recurring, or seasonal. Sampling-derived estimates, not exact counts.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues/{shortId}/history
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)
  • days number — How far back to read, 1–400. Defaults to 90.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_issue_history",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42,
      "days": 365
    }
  }
}

get_issue_activity

Activity feed for an issue: status changes, regressions, agent actions, with actor and time.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues/{shortId}/activity
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_issue_activity",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

resolve_issue

Mark an issue resolved. Requires event:write; audited.

• Scope: event:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/projects/{project}/issues/{shortId}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "resolve_issue",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

ignore_issue

Ignore an issue (mute alerts). Requires event:write; audited.

• Scope: event:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/projects/{project}/issues/{shortId}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "ignore_issue",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

unresolve_issue

Reopen an issue (back to unresolved). Requires event:write; audited.

• Scope: event:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/projects/{project}/issues/{shortId}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "unresolve_issue",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

bulk_update_issues

Set the status of up to 100 issues at once (resolve, ignore, or reopen). Requires event:write; audited.

• Scope: event:write · mutating, audited
• REST: POST /v1/orgs/{org}/projects/{project}/issues/bulk
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortIds array (required) — Issue short ids
  • status string (required) one of unresolved, resolved, ignored

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "bulk_update_issues",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortIds": [
        42,
        43
      ],
      "status": "resolved"
    }
  }
}

import_issues

Import a triaged issue backlog from another error tracker. Carries status, assignee and first/last seen; does NOT copy events — counts arrive as source metadata, never as live counters. DEFAULTS TO A DRY RUN: call once to see the coverage report, then again with dryRun=false. Export the issues from the old tracker and pass them here; Bugwatch never asks for a token to it. An issue that already exists here is left alone, because its triage was made in this product. Requires event:write.

• Scope: event:write · mutating, audited
• REST: POST /v1/orgs/{org}/projects/{project}/import-issues
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • source string (required) — Which tracker this came from, e.g. sentry or bugsnag
  • issues array (required) — The exported issues. Each needs at least an id; title, status, assignee (email), firstSeen, lastSeen, count, userCount, fingerprint and permalink are used when present.
  • dryRun boolean — Default true. Set false to actually create the issues.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "import_issues",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "source": "sentry",
      "issues": [
        {
          "id": "4512",
          "title": "TypeError: x is not a function",
          "status": "resolved"
        }
      ]
    }
  }
}

resolve_issues_from_commit

Resolve the issues a commit message closes. Reads GitHub-style trailers — Fixes #42, Closes COMPAT-42 — and resolves those issues, recording the commit on each one's activity feed. A bare mention without a closing verb resolves nothing. Call it after pushing a fix, or from CI. Requires event:write.

• Scope: event:write · mutating, audited
• REST: POST /v1/orgs/{org}/projects/{project}/resolve-by-commit
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • sha string (required) — Commit SHA
  • message string (required) — The full commit message, including its trailers
  • url string — Link to the commit

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "resolve_issues_from_commit",
    "arguments": {
      "org": "acme",
      "project": "web",
      "sha": "9f3a1c2",
      "message": "checkout: guard against a null cart\n\nFixes #412"
    }
  }
}

Events

list_events

List recent events for an issue (event id, timestamp, size), newest first; page with cursor. Use get_event for a body.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues/{shortId}/events
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)
  • cursor string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_events",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

get_event

Full processed event body (JSON, often 10–40 KB): exception with all frames, breadcrumbs, contexts, tags, request, sdk. Prefer get_fix_context unless you need everything.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues/{shortId}/events/{eventId}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)
  • eventId string (required) — 32-hex event id from list_events

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_event",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42,
      "eventId": "0123456789abcdef0123456789abcdef"
    }
  }
}

list_issue_attachments

Files the SDK attached to this issue's events (screenshots, logs, view hierarchies): filename, type, size, event id, newest first. Metadata only; the bytes are a binary download from GET …/attachments/{id}, not a tool result.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/issues/{shortId}/attachments
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • shortId integer (required) — Issue short id (the number in the dashboard URL)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_issue_attachments",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "shortId": 42
    }
  }
}

replay_raw_events

Re-process raw events already stored for a project over a time window — for after a pipeline incident. DEFAULTS TO A DRY RUN: call it once to see what it would cover, then again with dryRun=false. Events keep their original ids, so anything the 24h dedup window still remembers is a no-op; a window reaching further back is refused unless acceptDoubleCount is set, because those events would be counted and billed a second time. Transactions are never replayed. Requires event:write.

• Scope: event:write · mutating, audited
• REST: POST /v1/orgs/{org}/projects/{project}/replay
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • from number (required) — Window start, epoch ms
  • to number (required) — Window end, epoch ms
  • dryRun boolean — Default true. Set false to actually re-enqueue.
  • acceptDoubleCount boolean — Required to replay past the 24h dedup window; acknowledges that events will be re-counted and re-billed.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "replay_raw_events",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "from": 1757930000000,
      "to": 1757940000000
    }
  }
}

Performance

query_performance

Per-transaction performance overview for the last N hours: throughput, p50/p75/p95/p99 ms, failures. Sampling-aware.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/transactions
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • since integer — Window in hours
  • limit integer

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "query_performance",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "since": 24
    }
  }
}

get_transaction

One transaction in detail: throughput, p50/p75/p95/p99, failure rate, latency over time, the spans that own the most of its time, and the error issues seen in its recent traces (linked by trace id). Sampling-aware.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/transaction
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • name string — Exact transaction name, e.g. 'GET /api/orders'
  • since integer — Window in hours
  • interval integer — Series bucket size in minutes one of 1, 5, 15, 60, 1440
  • spans integer

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_transaction",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "name": "GET /api/orders",
      "since": 24
    }
  }
}

get_web_vitals

p75 Core Web Vitals per transaction for the last N hours (LCP, FCP, CLS, INP, TTFB), sampling-aware. A vital with no measurements is reported as null, never as zero.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/vitals
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • since integer — Window in hours
  • limit integer

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_web_vitals",
    "arguments": {
      "org": "acme",
      "project": "web-app",
      "since": 24
    }
  }
}

get_stats

Time series of event volume (errors, transactions, or sessions) for the last N hours at a chosen interval. Sampling-aware estimates.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/stats
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • dataset string one of errors, transactions, sessions
  • interval integer — Bucket size in minutes one of 1, 5, 15, 60, 1440
  • since integer — Window in hours
  • issue string — Restrict to one issue id (internal id, not short id)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_stats",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "dataset": "errors",
      "since": 24,
      "interval": 60
    }
  }
}

Releases

list_releases

Releases in the organization: version, status, commit, deploy count, last deploy, artifact bundles.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/releases
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_releases",
    "arguments": {
      "org": "acme"
    }
  }
}

create_release

Register a release version (optionally with commit SHA and URL). Requires release:write.

• Scope: release:write · mutating, audited
• REST: POST /v1/orgs/{org}/releases
• Arguments:
  • org string (required) — Organization slug
  • version string (required)
  • commitSha string
  • url string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_release",
    "arguments": {
      "org": "acme",
      "version": "1.4.2",
      "commitSha": "a3f9c2e"
    }
  }
}

create_deploy

Record a deploy of a release to an environment. Requires release:write.

• Scope: release:write · mutating, audited
• REST: POST /v1/orgs/{org}/releases/{version}/deploys
• Arguments:
  • org string (required) — Organization slug
  • version string (required)
  • environment string (required) — e.g. production
  • name string
  • url string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_deploy",
    "arguments": {
      "org": "acme",
      "version": "1.4.2",
      "environment": "production"
    }
  }
}

get_release_health

Session health of a release in a project: sessions, crashed/errored/abnormal, crash-free rate.

• Scope: event:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/releases/{version}/health
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • version string (required)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_release_health",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "version": "1.4.2"
    }
  }
}

Alerts

list_channels

Notification channels in the organization (Slack, webhook, email, PagerDuty) with masked targets.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/channels
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_channels",
    "arguments": {
      "org": "acme"
    }
  }
}

create_channel

Create a notification channel. config is one of: {kind:'slack_webhook',url}, {kind:'discord',url}, {kind:'teams',url}, {kind:'webhook',url}, {kind:'email',to}, {kind:'pagerduty',routingKey}, {kind:'voice',to}. Each url is host-checked: Slack on hooks.slack.com, Discord on discord.com/api/webhooks, Teams on a .logic.azure.com Workflows URL; anything else uses 'webhook'. voice takes an E.164 number. Stored encrypted. A 'webhook' channel gets its own signing secret (signingSecret, shown once in this response): its receiver verifies x-bugwatch-signature with it. An 'email' channel sends nothing until its address follows the confirmation link mailed to it (the response says verified:false until then). Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/channels
• Arguments:
  • org string (required) — Organization slug
  • name string (required)
  • config object (required) — Channel config (discriminated by kind)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_channel",
    "arguments": {
      "org": "acme",
      "name": "#alerts",
      "config": {
        "kind": "slack_webhook",
        "url": "https://hooks.slack.com/services/…"
      }
    }
  }
}

rotate_channel_secret

Give a webhook channel a new signing secret, returned once; the old one stops verifying at the next delivery. Also how a channel listed with legacySigning:true (signed with the deployment's shared secret) gets a secret of its own. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/channels/{id}/rotate-secret
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Channel id (a webhook channel)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "rotate_channel_secret",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

delete_channel

Delete a notification channel by id. Anything still pointing at it — services, monitors, alert rules — is detached in the same operation, so nothing is left holding a reference that pages nobody. A service left with no channels stops paging through channels entirely. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/channels/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Channel id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_channel",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

list_alert_rules

Alert rules for a project: conditions (new_issue, regression, frequency), filters, channel actions, throttle, enabled.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/alert-rules
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_alert_rules",
    "arguments": {
      "org": "acme",
      "project": "checkout-api"
    }
  }
}

create_alert_rule

Create an alert rule. conditions: [{type:'new_issue'}|{type:'regression'}|{type:'frequency',count,windowMinutes}]; filters: [{type:'environment'|'min_level'|'release',value}]; actions: [{type:'channel',channelId}]. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/projects/{project}/alert-rules
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • name string (required)
  • conditions array (required)
  • filters array
  • actions array (required)
  • environment string
  • frequencyS integer — Min seconds between notifications per issue
  • enabled boolean

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_alert_rule",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "name": "New issues → Slack",
      "conditions": [
        {
          "type": "new_issue"
        }
      ],
      "actions": [
        {
          "type": "channel",
          "channelId": "01J…"
        }
      ]
    }
  }
}

update_alert_rule

Replace an alert rule's definition (PATCH has replace semantics: send the full rule — name, conditions, actions — plus any of filters, environment, frequencyS, enabled). Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/projects/{project}/alert-rules/{id}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • id string (required) — Rule id
  • name string (required)
  • conditions array (required)
  • filters array
  • actions array (required)
  • environment string
  • frequencyS integer
  • enabled boolean

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "update_alert_rule",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "id": "01J…",
      "name": "New issues → Slack",
      "conditions": [
        {
          "type": "new_issue"
        }
      ],
      "actions": [
        {
          "type": "channel",
          "channelId": "01J…"
        }
      ],
      "enabled": false
    }
  }
}

delete_alert_rule

Delete an alert rule by id. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/projects/{project}/alert-rules/{id}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • id string (required)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_alert_rule",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "id": "01J…"
    }
  }
}

Uptime monitors

list_monitors

Uptime monitors in the organization: state (up/degraded/down/paused/pending), type, target, last check and latency.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/monitors
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_monitors",
    "arguments": {
      "org": "acme"
    }
  }
}

get_monitor

One uptime monitor: full check configuration (secrets redacted), interval, regions, thresholds, quorum, channels, current state and — for heartbeat monitors — the ping URL.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/monitors/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_monitor",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

create_monitor

Create an uptime monitor. check is one of: {type:'http',url,method?,expectedStatus?,keyword?,keywordAbsent?,timeoutMs?,maxLatencyMs?}; {type:'tcp',host,port}; {type:'tls',host,port?}; {type:'dns',name,recordType?,expected?}; {type:'heartbeat',expectedEverySeconds,graceSeconds?} (returns a ping URL the job must call). intervalSeconds must be at or above the plan minimum; regions are Cloudflare location hints (wnam, enam, weur, eeur, apac, oc, afr, sam) and a DOWN needs quorum of them to agree. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/monitors
• Arguments:
  • org string (required) — Organization slug
  • name string (required) — Display name
  • check object (required) — Check configuration, discriminated by type (http | tcp | tls | dns | heartbeat)
  • intervalSeconds integer — Seconds between checks
  • regions array — Probe regions
  • failureThreshold integer — Consecutive failures in a region before it counts as failing
  • recoveryThreshold integer — Consecutive passes before a region counts as passing
  • quorum integer — Regions that must agree before DOWN/UP (default ceil(regions/2))
  • channelIds array — Notification channel ids to page (see list_channels). With a serviceId set these are only the fallback, used if the escalation cannot start
  • serviceId string — Escalate through this service instead of paging channels: a DOWN opens an incident and its policy pages whoever is on call (see list_services)
  • enabled boolean

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_monitor",
    "arguments": {
      "org": "acme",
      "name": "Checkout API",
      "check": {
        "type": "http",
        "url": "https://api.example.com/health",
        "keyword": "ok"
      },
      "intervalSeconds": 60,
      "regions": [
        "wnam",
        "weur",
        "apac"
      ]
    }
  }
}

update_monitor

Update an uptime monitor: send only the fields to change (name, check, intervalSeconds, regions, failureThreshold, recoveryThreshold, quorum, channelIds, serviceId, enabled). A replaced check replaces the whole object. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/monitors/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id
  • name string
  • check object — Replacement check configuration (same shapes as create_monitor)
  • intervalSeconds integer
  • regions array
  • failureThreshold integer
  • recoveryThreshold integer
  • quorum integer
  • channelIds array
  • serviceId string — Escalate through this service; null detaches it
  • enabled boolean

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "update_monitor",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "intervalSeconds": 30,
      "channelIds": [
        "01J…"
      ]
    }
  }
}

delete_monitor

Delete an uptime monitor and stop its probes. History stays queryable until it ages out. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/monitors/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_monitor",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

pause_monitor

Pause an uptime monitor: probes stop and it never pages until resumed (use this during planned maintenance). Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/monitors/{id}/pause
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "pause_monitor",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

resume_monitor

Resume a paused uptime monitor. It restarts in the pending state and pages again once the consensus is DOWN. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/monitors/{id}/resume
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "resume_monitor",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

snooze_monitor

Quiet a monitor's paging for 15, 60, 240 or 1440 minutes while somebody works on it (#209). Checks, state and status pages carry on; only the DOWN page is withheld, and a monitor still down when the snooze ends is paged then. Never indefinite. To stop an open incident paging, acknowledge it instead. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/monitors/{id}/snooze
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id
  • minutes integer (required) — How long, in minutes one of 15, 60, 240, 1440

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "snooze_monitor",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "minutes": 60
    }
  }
}

unsnooze_monitor

Lift a monitor's snooze now. If it is down, the withheld page goes out at its next check. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/monitors/{id}/snooze
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "unsnooze_monitor",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

get_monitor_history

Uptime and latency history for one monitor: per-bucket uptime % with p50/p95, plus a per-region summary over the window. Sampling-aware estimates.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/monitors/{id}/history
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id
  • since integer — Window in hours
  • interval integer — Bucket size in minutes one of 1, 5, 15, 60, 1440

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_monitor_history",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "since": 24,
      "interval": 5
    }
  }
}

list_monitor_events

State-change history for one monitor (most recent 100): when it went up, degraded, down or paused, how many regions were failing, and why.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/monitors/{id}/events
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Monitor id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_monitor_events",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

On-call schedules

list_schedules

On-call schedules in the organization: name, timezone and how many rotation layers each has.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/schedules
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_schedules",
    "arguments": {
      "org": "acme"
    }
  }
}

get_schedule

One on-call schedule in full: timezone, every rotation layer (type, handoff wall-clock time, rotation order, restrictions, active dates) and every override.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/schedules/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_schedule",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

who_is_on_call

Who is on call for a schedule at an instant (default: now). Returns the user, whether an override or a rotation layer decided it, and null when the window is UNCOVERED — a gap is a real answer, not an error.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/schedules/{id}/on-call
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id
  • at integer — Epoch ms; defaults to now

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "who_is_on_call",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

get_schedule_coverage

Coverage for a schedule over a window (default: the next 7 days, 90 days maximum) as contiguous segments. Uncovered windows appear as segments with a null user — use this to find gaps before they page nobody.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/schedules/{id}/coverage
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id
  • from integer — Window start, epoch ms; defaults to now
  • to integer — Window end, epoch ms; defaults to 7 days after from

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_schedule_coverage",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

create_schedule

Create an on-call schedule. timeZone is an IANA zone name (never a UTC offset — offsets change, zones don't) and handoffs are wall-clock times in it, so a Monday 09:00 handoff stays 09:00 across DST. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/schedules
• Arguments:
  • org string (required) — Organization slug
  • name string (required)
  • timeZone string (required) — IANA zone, e.g. America/New_York
  • teamId string
  • layers array (required) — Rotation layers, highest position wins. Each: {name, position, rotationType: daily|weekly|custom, turnLengthSeconds (custom), handoffLocalTime like "09:00" (WALL CLOCK in the schedule timezone), handoffWeekday (weekly), startDate/endDate epoch ms, users (rotation order), restrictions [{kind: daily|weekly, startLocalTime, durationSeconds, weekday}].

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_schedule",
    "arguments": {
      "org": "acme",
      "name": "Platform primary",
      "timeZone": "America/New_York",
      "layers": [
        {
          "name": "Weekly primary",
          "position": 1,
          "rotationType": "weekly",
          "handoffLocalTime": "09:00",
          "handoffWeekday": 1,
          "startDate": 1767225600000,
          "users": [
            "usr_ada",
            "usr_bo",
            "usr_cy"
          ],
          "restrictions": []
        }
      ]
    }
  }
}

update_schedule

Update a schedule's name, timezone, team or layers. Sending layers REPLACES the whole stack — positions, rotation order and restrictions only make sense as a set. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/schedules/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id
  • name string
  • timeZone string
  • teamId string
  • layers array — Rotation layers, highest position wins. Each: {name, position, rotationType: daily|weekly|custom, turnLengthSeconds (custom), handoffLocalTime like "09:00" (WALL CLOCK in the schedule timezone), handoffWeekday (weekly), startDate/endDate epoch ms, users (rotation order), restrictions [{kind: daily|weekly, startLocalTime, durationSeconds, weekday}].

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "update_schedule",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "timeZone": "Europe/London"
    }
  }
}

delete_schedule

Delete an on-call schedule. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/schedules/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_schedule",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

create_override

Put someone on call for a window regardless of the rotation — a cover for illness, holiday or a handover. Overrides beat every layer. Half-open: [startsAt, endsAt). Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/schedules/{id}/overrides
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id
  • userId string (required) — User taking the shift
  • startsAt integer (required) — Epoch ms, inclusive
  • endsAt integer (required) — Epoch ms, exclusive

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_override",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "userId": "usr_sam",
      "startsAt": 1767225600000,
      "endsAt": 1767268800000
    }
  }
}

delete_override

Remove an override, handing the window back to the rotation. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/schedules/{id}/overrides/{overrideId}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id
  • overrideId string (required) — Override id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_override",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "overrideId": "01J…"
    }
  }
}

list_devices

The phones registered to you in this organization, with when each last confirmed a delivery. Push tokens are never returned — this is for recognising a device, not for reading its credentials. A device that stopped confirming deliveries is going stale and stops being counted on for pages.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/devices
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_devices",
    "arguments": {
      "org": "acme"
    }
  }
}

revoke_device

Stop paging one of your devices and clear its push credentials — a lost or replaced phone. Yours only; revoking somebody else's device would take them off call without telling them. Requires a session.

• Scope: org:read · mutating, audited
• REST: DELETE /v1/orgs/{org}/devices/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Device id from list_devices

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "revoke_device",
    "arguments": {
      "org": "acme",
      "id": "01JDEVICE"
    }
  }
}

get_notification_rules

Your notification rules for this organization: which categories are on, quiet hours, active mutes and which devices take pages. Pages are not among the things that can be switched off — an escalation targeting you always rings.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/notification-prefs
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_notification_rules",
    "arguments": {
      "org": "acme"
    }
  }
}

set_notification_rules

Replace your notification rules for this organization. Send the whole object. categories may not switch page off, and every mute needs an expiry (at most 30 days) — an indefinite silent mute is how a page gets missed. Requires a session.

• Scope: org:read · mutating, audited
• REST: PUT /v1/orgs/{org}/notification-prefs
• Arguments:
  • org string (required) — Organization slug
  • categories object — Category → on/off. page cannot be false.
  • quietHours object — {startLocalTime:"22:00", endLocalTime:"08:00", timeZone:"Europe/London"} or null. Never applies to pages.
  • mutes array — [{scope:{kind:"all_non_page"|"category"|"thread", …}, until: epochMs}]
  • pageDeviceIds array — Devices that take pages; empty means all of them

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "set_notification_rules",
    "arguments": {
      "org": "acme",
      "quietHours": {
        "startLocalTime": "22:00",
        "endLocalTime": "08:00",
        "timeZone": "Europe/London"
      }
    }
  }
}

diff_imported_schedule

Compare an Opsgenie or PagerDuty schedule against one of ours over a 30-day window, without importing anything. Pass the provider's schedule payload as source, name the provider, and give a userMap from their usernames to our user ids. Returns what the mapping could not carry across, then the coverage differences. Read-only: it stores nothing and cannot activate a rotation — a migration needs human sign-off before it pages anybody.

• Scope: org:read · read-only
• REST: POST /v1/orgs/{org}/schedules/{id}/import-diff
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Schedule id
  • provider string — Which system the payload came from. Default opsgenie. one of opsgenie, pagerduty
  • source object (required) — The provider's schedule payload — Opsgenie (id, timezone, rotations[]) or PagerDuty (id, time_zone, schedule_layers[], overrides[])
  • userMap object — Their username or id → our user id. Unmapped people are reported, never assumed.
  • from integer — Window start, epoch ms. Defaults to now.
  • days integer — Window length in days (1–30, default 30)
  • boundaryToleranceMs integer — Differences shorter than this are listed separately as boundary noise. Never reduces uncovered time.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "diff_imported_schedule",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "source": {
        "id": "sch-og",
        "timezone": "Europe/London",
        "rotations": []
      },
      "userMap": {
        "ada@acme.test": "usr_ada"
      }
    }
  }
}

Incidents & escalation

list_incidents

Incidents in the organization, newest first. Filter by status (triggered, acknowledged, resolved) — triggered is the list that matters during an outage: nobody has picked those up yet.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/incidents
• Arguments:
  • org string (required) — Organization slug
  • status string one of triggered, acknowledged, resolved
  • limit integer

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_incidents",
    "arguments": {
      "org": "acme",
      "status": "triggered"
    }
  }
}

get_incident

One incident with its full timeline: when it triggered, every escalation step, who was paged on which channel, who acknowledged and when it resolved. Read this before asking anyone what happened.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/incidents/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Incident id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_incident",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

trigger_incident

Open an incident and start its escalation chain. A repeated call with the same dedupKey joins the open incident instead of paging a second time. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/incidents
• Arguments:
  • org string (required) — Organization slug
  • serviceId string (required) — Service id
  • title string (required)
  • summary string
  • severity string one of sev1, sev2, sev3, sev4
  • urgency string one of low, high
  • dedupKey string — Repeated triggers with this key join the open incident

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "trigger_incident",
    "arguments": {
      "org": "acme",
      "serviceId": "01J…",
      "title": "Checkout API returning 500s",
      "severity": "sev2"
    }
  }
}

acknowledge_incident

Acknowledge an incident: escalation stops immediately. If the service sets an ack timeout and nobody resolves it, it re-escalates — acking is not a way to make it go away. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/incidents/{id}/acknowledge
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Incident id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "acknowledge_incident",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

resolve_incident

Resolve an incident: escalation stops and the timer clears. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/incidents/{id}/resolve
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Incident id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "resolve_incident",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

list_escalation_policies

Escalation policies: how many rules each chain has, whether it repeats, and its fallback target.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/escalation-policies
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_escalation_policies",
    "arguments": {
      "org": "acme"
    }
  }
}

get_escalation_policy

One escalation policy: every rule in order, its targets (schedule or user), delay and urgency.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/escalation-policies/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Escalation policy id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_escalation_policy",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

create_escalation_policy

Create an escalation policy. Rule 0 fires the moment an incident opens; each rule's delaySeconds is how long to wait before the NEXT one. Set fallbackUserId so an uncovered rotation still reaches somebody. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/escalation-policies
• Arguments:
  • org string (required) — Organization slug
  • name string (required)
  • rules array (required) — Ordered escalation rules. Each: {position (0 fires immediately), delaySeconds (wait before the NEXT rule), urgency: low|high, targets: [{kind: schedule|user, targetId}]}. Low urgency notifies without paging.
  • repeatCount integer — Times to run the whole chain again
  • repeatDelaySeconds integer
  • fallbackUserId string — Paged when a rule resolves to nobody
  • teamId string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_escalation_policy",
    "arguments": {
      "org": "acme",
      "name": "Platform chain",
      "fallbackUserId": "usr_lead",
      "rules": [
        {
          "position": 0,
          "delaySeconds": 300,
          "urgency": "high",
          "targets": [
            {
              "kind": "schedule",
              "targetId": "01J…"
            }
          ]
        },
        {
          "position": 1,
          "delaySeconds": 300,
          "urgency": "high",
          "targets": [
            {
              "kind": "user",
              "targetId": "usr_lead"
            }
          ]
        }
      ]
    }
  }
}

update_escalation_policy

Update a policy's name, repeat behaviour, fallback or rules. Sending rules REPLACES the whole chain — positions and targets only mean anything as an ordered set. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/escalation-policies/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Escalation policy id
  • name string
  • rules array — Ordered escalation rules. Each: {position (0 fires immediately), delaySeconds (wait before the NEXT rule), urgency: low|high, targets: [{kind: schedule|user, targetId}]}. Low urgency notifies without paging.
  • repeatCount integer
  • repeatDelaySeconds integer
  • fallbackUserId string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "update_escalation_policy",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "repeatCount": 2
    }
  }
}

delete_escalation_policy

Delete an escalation policy. Services pointing at it stop paging, so repoint them first. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/escalation-policies/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Escalation policy id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_escalation_policy",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

list_services

Services: what an incident is about, which escalation policy it pages, which channels it notifies, and its auto-resolve and ack-timeout settings.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/services
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_services",
    "arguments": {
      "org": "acme"
    }
  }
}

create_service

Create a service: a name, the escalation policy its incidents page, the channels they notify, and optionally the product-line-1 project it belongs to. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/services
• Arguments:
  • org string (required) — Organization slug
  • name string (required)
  • escalationPolicyId string — Escalation policy id
  • projectId string — Links the service to a Bugwatch project
  • channelIds array — Notification channels to page
  • autoResolveSeconds integer
  • ackTimeoutSeconds integer — Re-escalate an ack nobody followed up

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_service",
    "arguments": {
      "org": "acme",
      "name": "Checkout",
      "escalationPolicyId": "01J…",
      "channelIds": [
        "01J…"
      ],
      "ackTimeoutSeconds": 1800
    }
  }
}

update_service

Update a service's name, policy, channels, project link, auto-resolve or ack timeout. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/services/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Service id
  • name string
  • escalationPolicyId string — Escalation policy id
  • projectId string
  • channelIds array
  • serviceId string — Escalate through this service; null detaches it
  • autoResolveSeconds integer
  • ackTimeoutSeconds integer

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "update_service",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "ackTimeoutSeconds": 900
    }
  }
}

delete_service

Delete a service. Its open incidents keep their escalation; new alerts for it stop. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/services/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Service id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_service",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

get_incident_context

What else changed around an incident: error rate and probe latency before vs after it opened, and deploys in the two hours before. Ranked evidence for a responder who has already been paged — it never delays or replaces a page, and cannot acknowledge, resolve or suppress anything. Says plainly when nothing stood out, and which checks it was able to run.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/incidents/{id}/context
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Incident id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_incident_context",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

draft_incident_retro

A retro draft assembled from the incident's own timeline: how long detection, acknowledgement and the fix each took, what the response itself got wrong (pages that were never delivered, a chain that reached its fallback, nobody answering), and the questions left to fill in. It measures; it does not write the analysis. Read-only.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/incidents/{id}/retro
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Incident id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "draft_incident_retro",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

get_incident_brief

Catch up on an incident that is still happening: whether anybody has it, who has been paged and who answered, what has already been tried, and what is wrong with the RESPONSE itself — pages that were never delivered, a service with no escalation policy, a source that fired again. Written for somebody who was just paged and has ten seconds. Distinguishes 'nobody answered' from 'nothing was ever delivered', which have different next actions. Read-only.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/incidents/{id}/brief
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Incident id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_incident_brief",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

draft_status_update

A status-page update drafted for CUSTOMERS from what the incident records: the stage (investigating, identified, monitoring, resolved), which components are affected in the words the public page uses, how long it has been going on, and when the next update is promised. It never states a cause, never names anything internal, and refuses to write an all-clear while a component is still reporting a fault. Optional stage asks for monitoring; optional nextUpdateMinutes sets the cadence. Read-only — publishing is a human action.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/incidents/{id}/status-draft
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Incident id
  • stage string — Ask for a stage. Only monitoring is honoured, and only once the incident is acknowledged — every other stage is derived from the incident itself, because an operator who can choose the stage can choose resolved during an outage. one of investigating, identified, monitoring, resolved
  • nextUpdateMinutes number — Minutes until the next promised update. Defaults per stage.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "draft_status_update",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

Status pages

list_status_pages

Public status pages in the organization: slug, public URL, and whether each is published.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/status-pages
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_status_pages",
    "arguments": {
      "org": "acme"
    }
  }
}

get_status_page

One status page with its components, in display order, and what each component reads its health from.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/status-pages/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_status_page",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

create_status_page

Create a public status page. components are what readers see, in order — each one a public NAME for a monitor or service, so nobody outside learns your monitor ids. Pages are unpublished until published is true. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/status-pages
• Arguments:
  • org string (required) — Organization slug
  • slug string (required) — Public URL segment: lowercase letters, digits and dashes
  • name string (required) — Page title, e.g. "Acme Status"
  • headline string — One line shown above the components
  • published boolean — False (the default) authors the page without serving it
  • components array — What the page shows, in display order. Each: {name (the public label, e.g. "Checkout"), description?, sourceKind: monitor|service, sourceId}. The name is what readers see — a monitor id never reaches the page.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_status_page",
    "arguments": {
      "org": "acme",
      "slug": "acme",
      "name": "Acme Status",
      "published": true,
      "components": [
        {
          "name": "Checkout",
          "sourceKind": "monitor",
          "sourceId": "01J…"
        }
      ]
    }
  }
}

update_status_page

Update a status page. Sending components REPLACES the whole list — order is the display order, so it only means anything as a set. Setting published to false takes the page down immediately. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/status-pages/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id
  • slug string — Changing this changes the public URL and retires the old one
  • name string
  • headline string
  • published boolean
  • components array — What the page shows, in display order. Each: {name (the public label, e.g. "Checkout"), description?, sourceKind: monitor|service, sourceId}. The name is what readers see — a monitor id never reaches the page.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "update_status_page",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "published": true
    }
  }
}

delete_status_page

Delete a status page. Its public URL stops answering immediately. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/status-pages/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "delete_status_page",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

list_status_page_domains

Custom domains on a status page: the hostname, whether it is live, and the DNS records still waiting to be added. state is pending, active or failed — a domain serves traffic only when it is active.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/status-pages/{id}/domains
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_status_page_domains",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

add_status_page_domain

Point a hostname you own at a status page. Returns the DNS records to add; the domain goes live on its own once they resolve and the certificate issues — usually minutes. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/status-pages/{id}/domains
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id
  • hostname string (required) — The hostname to serve the page on, e.g. status.acme.com. No wildcards.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "add_status_page_domain",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "hostname": "status.acme.com"
    }
  }
}

verify_status_page_domain

Re-check a custom domain with the certificate authority now, instead of waiting for the next automatic check. Use after adding the DNS records. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/status-pages/{id}/domains/{domainId}/verify
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id
  • domainId string (required) — Custom domain id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "verify_status_page_domain",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "domainId": "01J…"
    }
  }
}

remove_status_page_domain

Stop serving a status page on a custom domain and release its certificate. The hostname stops answering immediately. Requires alert:write.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/status-pages/{id}/domains/{domainId}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id
  • domainId string (required) — Custom domain id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "remove_status_page_domain",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "domainId": "01J…"
    }
  }
}

list_maintenance

Scheduled maintenance windows on a status page, newest first, including past ones. A window with no components covers the whole page.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/status-pages/{id}/maintenance
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_maintenance",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

schedule_maintenance

Schedule a maintenance window on a status page. Components inside an open window read 'maintenance' — unless they are genuinely down, which always wins. Omit componentIds to cover the whole page. Times are epoch ms. Requires alert:write; audited.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/status-pages/{id}/maintenance
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id
  • title string (required) — What the notice says, e.g. "Database upgrade".
  • body string — Optional detail shown under the title.
  • startsAt number (required) — Epoch ms when the window opens.
  • endsAt number (required) — Epoch ms when it closes. Must be after startsAt.
  • componentIds array — Component ids this covers. Omit or leave empty for the whole page.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "schedule_maintenance",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "title": "Database upgrade",
      "startsAt": 1789000000000,
      "endsAt": 1789007200000
    }
  }
}

cancel_maintenance

Cancel a scheduled maintenance window. The status page stops showing it immediately. Requires alert:write; audited.

• Scope: alert:write · mutating, audited
• REST: DELETE /v1/orgs/{org}/status-pages/{id}/maintenance/{windowId}
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id
  • windowId string (required) — The window's id.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "cancel_maintenance",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "windowId": "01K…"
    }
  }
}

list_status_incidents

Incidents this status page is showing right now — open ones and any resolved in the last 7 days — each with the stage readers see and every update published about it. This is the page's own view, so it is what readers see and not a separate report of it.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/status-pages/{id}/incidents
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_status_incidents",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

post_status_update

Publish an update about an incident to a status page: the stage it is at (investigating, identified, monitoring, resolved) and what you want readers to know, in your words. This is PUBLIC the moment it is written — the page rebuilds immediately. The stage you set is what the page shows, overriding the one derived from whether anybody has acknowledged the page; a closed incident always reads resolved whatever was last published. Requires alert:write; audited.

• Scope: alert:write · mutating, audited
• REST: POST /v1/orgs/{org}/status-pages/{id}/incidents/{incidentId}/updates
• Arguments:
  • org string (required) — Organization slug
  • id string (required) — Status page id
  • incidentId string (required) — The incident to publish about. It must be one this page shows.
  • state string (required) — investigating: we know something is wrong. identified: we know what. monitoring: a fix is out and we are watching. resolved: it is over. one of investigating, identified, monitoring, resolved
  • body string (required) — What readers are told. Plain prose; no internal names or responder identities.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "post_status_update",
    "arguments": {
      "org": "acme",
      "id": "01J…",
      "incidentId": "01K…",
      "state": "identified",
      "body": "A bad deploy is returning errors on checkout. We are rolling it back."
    }
  }
}

DSN keys

list_keys

DSN keys for a project (public key, DSN, label, status). Use the DSN to configure an SDK.

• Scope: project:read · read-only
• REST: GET /v1/orgs/{org}/projects/{project}/keys
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_keys",
    "arguments": {
      "org": "acme",
      "project": "checkout-api"
    }
  }
}

create_key

Create an additional DSN key for a project (for rotation or a second environment). Requires project:write.

• Scope: project:write · mutating, audited
• REST: POST /v1/orgs/{org}/projects/{project}/keys
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • label string

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "create_key",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "label": "staging"
    }
  }
}

set_key_status

Enable or disable a DSN key. Disabled keys are rejected at ingest within seconds. Requires project:write.

• Scope: project:write · mutating, audited
• REST: PATCH /v1/orgs/{org}/projects/{project}/keys/{keyId}
• Arguments:
  • org string (required) — Organization slug
  • project string (required) — Project slug
  • keyId string (required)
  • status string (required) one of active, disabled

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "set_key_status",
    "arguments": {
      "org": "acme",
      "project": "checkout-api",
      "keyId": "01J…",
      "status": "disabled"
    }
  }
}

Members

list_members

Organization members with role (admin, member, viewer) and email.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/members
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_members",
    "arguments": {
      "org": "acme"
    }
  }
}

API tokens

list_tokens

API tokens in the organization (name, scopes, last used, expiry). Secrets are never returned. Requires org:admin.

• Scope: org:admin · read-only
• REST: GET /v1/orgs/{org}/tokens
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_tokens",
    "arguments": {
      "org": "acme"
    }
  }
}

revoke_token

Revoke an API token by id. Anything using it stops working immediately. Requires org:admin.

• Scope: org:admin · mutating, audited
• REST: DELETE /v1/orgs/{org}/tokens/{id}
• Arguments:
  • org string (required) — Organization slug
  • id string (required)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "revoke_token",
    "arguments": {
      "org": "acme",
      "id": "01J…"
    }
  }
}

Billing & usage

get_agent_usage

Fix-agent credits and usage: this month's allowance and what is left of it, purchased credit remaining, which models the plan may use and what they cost per million tokens, and the most recent runs.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/agent-usage
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_agent_usage",
    "arguments": {
      "org": "acme"
    }
  }
}

get_usage

Exact usage and plan for the organization: plan limits, overage flag, daily accepted events by category, credit balance and expiry.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/usage
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_usage",
    "arguments": {
      "org": "acme"
    }
  }
}

get_billing_plans

Plan catalog with prices, included events, retention, overage rate, and whether checkout is available.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/billing/plans
• Arguments:
  • org string (required) — Organization slug

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_billing_plans",
    "arguments": {
      "org": "acme"
    }
  }
}

set_overage

Turn pay-as-you-go overage on or off (off = hard cap at the plan's included events). Requires org:admin.

• Scope: org:admin · mutating, audited
• REST: POST /v1/orgs/{org}/billing/overage
• Arguments:
  • org string (required) — Organization slug
  • enabled boolean (required)

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "set_overage",
    "arguments": {
      "org": "acme",
      "enabled": true
    }
  }
}

Agents

list_agent_activity

Recent agent (MCP) actions across the organization: tool, arguments, token, outcome, time.

• Scope: org:read · read-only
• REST: GET /v1/orgs/{org}/agent-activity
• Arguments:
  • org string (required) — Organization slug
  • limit integer

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_agent_activity",
    "arguments": {
      "org": "acme"
    }
  }
}

Routes without a tool

• POST /v1/orgs/{org}/tokens — Agents must not mint credentials; tokens are created by a human in Settings → Agents / API tokens.
• POST /v1/orgs/{org}/members — Adding a member (or changing a role) grants access that outlives the credential that granted it, so a leaked token or a misled agent could keep its foothold after revocation (#276). Human-only, in Settings → Members.
• GET /v1/orgs/{org}/projects/{project}/attachments/{id} — The bytes of a file an SDK attached to an event: binary and always served as a download, not something a tool result can carry. Agents get the metadata from list_issue_attachments; the file is fetched from this route with an event:read token, or from the issue page (#240).
• GET /v1/orgs/{org}/export — A bulk download of every row the org owns, streamed as NDJSON for a backup or a move; far beyond what a tool result can carry. Use curl with an org:admin token (P2.2).
• GET /v1/orgs/{org}/projects/{project}/export/events — A bulk stream of a project's stored event bodies from R2; a download, not a tool result (P2.2).
• DELETE /v1/orgs/{org} — Deleting an organization erases all of its data and cannot be undone. Owner-only, signed-in-only, with the slug typed back, in Settings → General (P2.2).
• DELETE /v1/orgs/{org}/members/{userId} — A leaked token must not be able to remove the admins who would revoke it (#276). Human-only, in Settings → Members.
• POST /v1/orgs/{org}/billing/checkout — Browser redirect to Stripe Checkout; purchases stay human-initiated.
• POST /v1/orgs/{org}/billing/portal — Browser redirect to the Stripe Billing Portal.
• POST /v1/orgs/{org}/billing/agent-credits/checkout — Browser redirect to Stripe Checkout for a fix-agent credit pack; purchases stay human-initiated.
• POST /v1/orgs/{org}/projects/{project}/artifact-bundles — Binary source-map upload; use sentry-cli or the REST endpoint directly.
• POST /v1/orgs/{org}/repo-connection — Stores a credential that can write to the customer's source code; connecting a repository stays human-initiated, like minting a token.
• DELETE /v1/orgs/{org}/repo-connection — The same human-only gate as connecting: an agent must not be able to sever, or silently re-point, the org's source-code access.
• POST /v1/orgs/{org}/projects/{project}/issues/{shortId}/fix-agent — Server-sent event stream that runs a model; an agent wanting the same material should call get_fix_context and do its own reasoning rather than nest an agent inside a tool call.
• POST /v1/orgs/{org}/devices — A device registers itself with its own push credentials from the app; there is nothing for an agent to do here, and a tool that could write push tokens could redirect somebody's pages to another handset.
• POST /v1/orgs/{org}/chat — The mobile companion's own model loop, which runs THESE tools. An agent calling it would be nesting an agent inside a tool call to reach the same registry it already has — the same reason the fix-agent stream is exempt. Its mutations are the mobile allowlist and each needs a person to tap a confirmation card, which a tool call cannot do.
• POST /v1/orgs/{org}/test-page — Rings the caller's own physical phone at critical-alert insistence to prove their paging setup works. Session-only by construction, and making somebody's handset ring is a thing a person asks for about their own hardware, not a thing an agent does on their behalf.
• GET /v1/orgs/{org}/devices/vapid-key — The application server key a BROWSER needs to call PushManager.subscribe(). It is public key material for a client-side API an agent has no way to call; list_devices already reports what is registered.
• POST /v1/orgs/{org}/services/{id}/inbound-key — Mints a credential that can open an incident on this service, shown exactly once. Minting stays human-initiated for the same reason API tokens do — and rotating one silently breaks every integration still pointed at the old key, which is a decision a person makes.
• POST /v1/orgs/{org}/devices/{id}/receipt — A handset confirms a push it actually received. An agent confirming on its behalf would mark a silent device fresh, which is exactly the state the receipt exists to detect.