opsgenie-mcp

MCP server for Opsgenie (Atlassian's incident/alert management and on-call platform) — exposes the full public Opsgenie REST API as MCP tools.

⚠️ Vendor note: Atlassian has announced Opsgenie is reaching end of support and recommends migrating to Jira Service Management or Compass. The public API documented here (docs.opsgenie.com) is still live and unchanged as of this build, but expect it to eventually be deprecated.

Overview

• Stateless HTTP service. No credentials are ever persisted — each request supplies its own API key via a header, used only for the lifetime of that single request.

• Supports concurrent requests; per-request credential isolation is done via Python contextvars, not a global/shared client instance.

• Entry points: POST /mcp (MCP protocol) and GET /health (health check).

• Default port: 8080 (configurable via MCP_HTTP_PORT).

Related MCP server: mcp-server-splunk-oncall

Scope

22 tools, trimmed down from an original 213-tool full-API build (2026-08-04), all in the single alert category. MSPbots' own stored integration config for this vendor calls exactly 1 endpoint (GET /v2/alerts, read-only, → opsgenie_alert_list_alerts). Per the "actual usage + same-category core CRUD" scope decision, every other category (Incident, Incident Timeline, Integration, Heartbeat, Alert & Notification Policy, Policy (legacy v1), Maintenance, Account, User, Custom User Role, Contact, Notification Rule (+ Step), Team (+ Member/ Role/Routing Rule), Schedule (+ Rotation/Override), Escalation, Who Is On Call, Forwarding Rule, Service (+ Incident Rules/Templates), Incident Templates — 27 categories, ~178 tools) was removed entirely, since MSPbots doesn't touch any of them; within the alert category itself, 35 original tools were cut to the 22 core alert-lifecycle operations (list/get/create/close/delete/acknowledge/unacknowledge/note/tags/ assign/responder/team/escalate/snooze/update message-description-priority/ count/logs/request-status) — dropped were the saved-search sub-feature (5 tools), the attachment sub-feature (4 tools), custom-properties add/remove (2 tools), execute_custom_action (1), and list_alert_recipients (1), none of which are core alert CRUD.

Source data for the kept tools was originally extracted by fetching docs.opsgenie.com/docs/alert-api and structuring each documented operation (method, path, path/query params, JSON body fields) into a machine-readable format — the same codegen-from-structured-spec approach used for other large-API vendors in this program (ConnectSecure, Dynu, Jira Data Center), adapted here because Opsgenie has no downloadable OpenAPI/Postman spec, only prose documentation. If a removed category is needed later, the same source pages (docs.opsgenie.com/docs/*-api) can be re-parsed the same way.

Authentication

Opsgenie uses a static API key tied to a specific Opsgenie integration (created in Opsgenie's Settings → Integrations → API). MSPbots' own integration convention sends this key as Authorization: GenieKey <apiKey>, matching Opsgenie's own documented format, and this server forwards it exactly that way.

HEADER 授权参数说明

Missing the required header returns 401:

{
  "error": "Missing credentials",
  "message": "This server requires the X-Opsgenie-Api-Key header",
  "required_headers": ["X-Opsgenie-Api-Key"],
  "optional_headers": ["X-Opsgenie-Base-Url"]
}

Environment Variables

MCP Endpoint

• POST /mcp — MCP protocol (streamable HTTP transport)

• GET /health — health check, returns {"status": "ok", "service": "opsgenie-mcp", "transport": "http"}

Tool List

Tool names are opsgenie_<category>_<operation>, derived from each operation's heading in the official docs (e.g. "List Alerts" in the alert category → opsgenie_alert_list_alerts). body parameters are accepted as a generic dict — the exact field list for each is documented in that tool's own docstring (extracted from the source docs), and the full field schema is available in Opsgenie's own API reference (linked below).

测试示例

# Health check
curl -s http://localhost:8080/health

# Call a tool via the MCP protocol (streamable HTTP) — requires an
# initialize handshake first per the MCP spec; abbreviated example below
# shows the tool-call request body only:
curl -s -X POST http://localhost:8080/mcp \
  -H "X-Opsgenie-Api-Key: <your-opsgenie-api-key>" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "mcp-session-id: <session-id-from-initialize>" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "opsgenie_alert_list_alerts",
      "arguments": {"limit": "5"}
    }
  }'

Live-verified (2026-07-30): opsgenie_alert_list_alerts (the same endpoint MSPbots itself calls) was called end-to-end through this running server with a real test API key and returned real alert data (message, status, priority, responders, integration, etc.) from the account. Two other tools in different categories (opsgenie_team_list_teams, opsgenie_account_get_account_info) were also called with the same key and correctly returned Opsgenie's own 403 API Key is not granted with configuration access — confirming the request/auth-header plumbing works correctly outside the alert category too; this specific test key is just scoped to alert-level access only (a normal Opsgenie integration-key restriction, not a bug).

API Reference

• Overview: https://docs.opsgenie.com/docs/api-overview

• Authentication: https://docs.opsgenie.com/docs/authentication

• Per-category docs are linked from the sidebar at https://docs.opsgenie.com/docs/alert-api

Known Gaps

• Trimmed from 213 to 22 tools on 2026-08-04. The original build covered the full public API across 28 categories per an earlier scope decision. A later scope decision cut this back to MSPbots' actually-used category (alert) plus its core CRUD — see the Scope section above for exactly what was kept/dropped within alert and the full list of the 27 removed categories (~178 tools). If a removed category is needed later, the same docs.opsgenie.com/docs/*-api pages can be re-parsed the same way the kept tools were generated.

• Several kept tools still mutate real Opsgenie alert state (opsgenie_alert_delete_alert, opsgenie_alert_close_alert, opsgenie_alert_create_alert, etc.) — treat these as irreversible/state-changing and confirm with a human before invoking.

• body parameters are untyped (dict) rather than fully modeled — each tool's docstring lists the field names extracted from Opsgenie's docs, but reproducing all of them as typed Python parameters was out of scope for a mechanically-generated server.

• Source data was extracted from prose documentation, not a machine spec — Opsgenie has no downloadable OpenAPI/Postman collection, so every operation's method/path/params was parsed from docs.opsgenie.com's HTML.

• Only opsgenie_alert_list_alerts (MSPbots' own endpoint) was live-verified with real data returned. The remaining 21 tools are structurally correct (schema validated, MCP-protocol tools/list confirmed, all pass compile()) but not individually smoke-tested — several are write/destructive operations that would create, modify, or close real alerts, so they weren't exercised against the live test account.

• Vendor is reaching end-of-support (see the warning banner at the top of this README) — Atlassian recommends migrating to Jira Service Management or Compass. This MCP targets the still-live Opsgenie public API as of this build.