About AlertOps
AlertOps is a read-only MCP server that exposes AlertOps incident alerting and on-call data to AI assistants and WYRE Conduit. Its tools cover only 11 GET operations and do not call endpoints that create or modify alerts.
Key features
- 01List and filter paginated alerts
- 02Look up users and contact methods
- 03List groups, topics, and integrations
- 04Check who is on call now
- 05Query user or group schedules
- 06Limit access to GET operations
Use cases
- 01Review and filter alerts
- 02Check user contact details and group membership
- 03Identify the current on-call person
- 04Inspect on-call schedules for a date range
MCP support
The MCP server provides clients with 11 read-only GET operations for AlertOps alerts, users, groups, topics, integrations, and schedules; it offers no write tools.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
docker run -i --rm -e ALERTOPS_API_KEY -e MCP_TRANSPORT -e AUTH_MODE -e LOG_LEVEL ghcr.io/wyre-ai/alertops-mcp:v1.0.2ALERTOPS_API_KEYSecretAlertOps API key, sent as the `APIKey` query parameter on every REST call. Generate one under Account Settings -> Configurations -> Integrations -> API.MCP_TRANSPORTTransport mode for the server. Set to 'stdio' for local CLI use; the image defaults to 'http' for gateway hosting.AUTH_MODESecretCredential source: 'env' reads vars locally, 'gateway' expects header injection from the WYRE MCP Gateway.LOG_LEVELLog verbosity: debug, info, warn, errorLocal servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
ghcr.io/wyre-ai/alertops-mcp:v1.0.2stdioTools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEAlertOps MCP Server
MCP server for AlertOps's incident alerting/on-call API - alerts, message topics, users and their contact methods, groups, configured inbound integrations, and on-call schedules, for AI assistants and the WYRE Conduit gateway.
Authentication
AlertOps' REST API (https://api.alertops.com/api/..., per its published OpenAPI spec at api.alertops.com/swagger/v1/swagger.json) authenticates with a static account-level API key, generated under Account Settings -> Configurations -> Integrations -> API. Unlike most WYRE Conduit sidecars, the key is not sent as a bearer token or custom auth header - AlertOps requires it as an APIKey query-string parameter on every single request, GET and POST alike (live-verified 2026-09-20: an invalid key returns a clean {"reason":"Invalid API Key"} HTTP 401 on every endpoint this connector calls). This connector receives the key per-request via a custom header (never OAuth, same shape as every other static-key WYRE Conduit sidecar) and itself appends ?APIKey=<key> when calling AlertOps' API. In gateway mode the key arrives per-request via the X-AlertOps-Api-Key header; in local/stdio mode it's read once from ALERTOPS_API_KEY. HTTP /mcp also requires an X-Gateway-S2S signature (see HTTP authentication below). The HTTP server refuses to start without a real CONDUIT_S2S_SECRET.
Two separate claims here, at deliberately different confidence levels - don't collapse them into one "read-only" statement:
- Structurally verified (checked directly, stated with full confidence): this connector's own code makes zero calls to a write/mutating AlertOps endpoint. Every function in
client.tscalls one of the 11 documented GET operations listed under Tools below; noPOSTcall exists anywhere insrc/. - Vendor-documented, not independently verified (hedged deliberately): AlertOps ships role-based access control with built-in and custom roles built from individually-grantable read-only entitlements (e.g.
Messages_View_GlobalAccess,User_View_GlobalAccess,Groups_View_GlobalAccess) - AlertOps' own docs describe a custom role built this way as letting a user "browse incidents, reports, users, and groups but cannot create, modify, or delete anything" (help.alertops.com, Role Based Access Control). A distinct User API Key, shown on a user's own Profile page, is described elsewhere in AlertOps' docs as gated by theUserAPIKey_Addentitlement, implying an API key issued this way inherits the issuing user's role/entitlements. Whether that's actually enforced server-side against this connector'sAPIKeyquery parameter and its write endpoints specifically - i.e. whether AlertOps' backend rejects a write call (Create Alert, Close Alert, Assign, etc.) made with a key belonging to a read-only-scoped user, versus that restriction only ever being surfaced in AlertOps' own UI - has not been tested by WYRE. Nobody sent a write call against a live key to check, correctly: that would risk creating a real page/notification against real on-call staff, not something to run without consent. Do not read this connector, or this README, as having established that a read-only-provisioned AlertOps API key technically cannot perform writes - only that this connector's own code never attempts one, and that AlertOps' documented role system supports provisioning a key that way.
Configuration
Copy .env.example. Leave CONDUIT_S2S_SECRET empty in that file.
| Env var | Description |
|---|---|
ALERTOPS_API_KEY | API key issued under Account Settings -> Configurations -> Integrations -> API. Used by stdio and by HTTP when AUTH_MODE=env. |
MCP_TRANSPORT | stdio (default) or http. |
AUTH_MODE | env (local default; reads ALERTOPS_API_KEY) or gateway (Docker image and Compose default). Gateway mode requires X-AlertOps-Api-Key on each /mcp request. |
CONDUIT_S2S_SECRET | Required to start the HTTP transport. Empty values and known placeholders (changeme, replace-with-a-real-secret, your-s2s-secret, and similar) are rejected: the process logs an error and exits non-zero. The value is never logged. Compose will not start when this is unset. |
MCP_ALLOW_INSECURE_DEV | Set to exactly 1 for local development only, to start HTTP when CONDUIT_S2S_SECRET is empty. Logs a warning, skips the X-Gateway-S2S check, and listens on 127.0.0.1 only. A non-loopback MCP_HTTP_HOST is ignored in that mode. A known placeholder secret is still rejected. |
MCP_HTTP_HOST | Bind address for HTTP. Defaults to 127.0.0.1 when unset. A configured CONDUIT_S2S_SECRET honors this value (the container image sets 0.0.0.0). |
MCP_HTTP_PORT | HTTP port. Defaults to 8080. |
LOG_LEVEL | debug | info (default) | warn | error. |
/health stays unauthenticated and does not read credentials. The stdio transport does not use CONDUIT_S2S_SECRET.
HTTP authentication
With MCP_TRANSPORT=http and CONDUIT_S2S_SECRET set, every /mcp request must include X-Gateway-S2S. The value is:
t=<unix seconds>,v1=<hex hmac-sha256>
v1 is HMAC-SHA256 over the ASCII message t=<unix seconds> (the same t= prefix that appears in the header), keyed by CONDUIT_S2S_SECRET, encoded as 64 lowercase hex characters. A missing or invalid header returns 401. The secret is never logged.
In AUTH_MODE=gateway, /mcp also requires X-AlertOps-Api-Key. A request missing that header returns 401. Gateway mode uses that header as the only AlertOps credential for the request.
Request limits
- HTTP request bodies are capped at 256 KiB (262144 bytes). After the
X-Gateway-S2Scheck passes, aContent-Lengthabove the cap returns 413 with a JSON-RPC error; bodies withoutContent-Lengthare capped by the MCP SDK transport. Unauthenticated requests still get 401 first. tools/callarguments are capped at 10,000 elements (every array item and object key at any depth). Larger inputs return anisErrortool result and are never dispatched.
These limits address CVE-2026-111768 / GHSA-46jr-4wx8-v453 (memory exhaustion during validation) and GHSA-f897-62hf-6gxq (no request body / batch cap).
Tools
11 read-only tools, one per GET operation in AlertOps' published REST API spec (25 operations total; the other 14 are POST/write and are deliberately excluded - see Scope below).
Alerts
alertops_list_alerts- list alerts, filterable by status/date range/group/integration/escalation, paginated.alertops_get_alert_id_by_inbound_message_id- resolve the AlertID an inbound integration's messageID was mapped to.alertops_get_alert_id_by_unique_inbound_id- resolve the AlertID an inbound integration's uniqueInboundID was mapped to.
Users
alertops_list_users- list users (username, full name, user type).alertops_get_user_contact_methods- list a user's configured contact methods (email, phone/SMS). Contact PII.
Groups
alertops_list_groups- list groups (group ID -> group name).
Topics
alertops_list_topics- list message topics used for alert routing/escalation.
Integrations
alertops_list_integrations- list configured inbound integrations (integration rule ID -> integration type).
Schedule
alertops_get_oncall_now- list who is on-call right now, for every group or one specific group.alertops_get_user_schedule- a user's on-call schedule for a date range (max 3 months).alertops_get_group_schedule- a group's on-call schedule for a date range (max 3 months).
Scope
This is a deliberately narrow, read-only v1 surface, hard-scoped to exactly 11 of AlertOps' 25 documented operations. AlertOps is a live incident-paging system - every excluded operation either mutates state or can trigger a real page/notification to a human, so none of it is implemented here, by design, not by oversight:
Hard-excluded (creates or mutates a live alert - can page a human) - never implemented:
POST /api/alert/Create(Alert_Create) - creates a new alert, which pages/notifies its recipients.POST /api/alert/Close(Alert_Close) - closes a live alert.POST /api/alert/Update(Alert_Update) - updates a field on a live alert.POST /api/alert/Assign(Alert_Assign) - assigns a user to an alert.POST /api/alert/Note(Alert_Note) - adds a note to an alert.POST /api/alert/AddRecipients(Alert_AddRecipients) - adds recipients to an alert, i.e. pages additional people.POST /api/alert/Reply(Alert_Reply) - replies to an alert.
Hard-excluded (provisioning/mutation - users, groups, schedules, integrations) - never implemented:
POST /api/user/CreateUser(User_CreateUser) - creates a new user.POST /api/user/CreateOutOfOffice(User_CreateOutOfOffice) - creates an out-of-office record, which changes who gets paged.POST /api/group/CreateGroup(Group_CreateGroup) - creates a new group.POST /api/group/AddGroupMember(Group_AddGroupMember) - adds a member to a group, which changes who gets paged.POST /api/schedule/CreateFixedSchedule(Schedule_CreateFixedSchedule) - creates a fixed on-call schedule.POST /api/schedule/CreateRecurringSchedule(Schedule_CreateRecurringSchedule) - creates a recurring on-call schedule.POST /api/integration/CreateMaintenanceWindows(Integration_CreateMaintenanceWindows) - creates a maintenance window, which suppresses alerting.
They can be added as a follow-up if there's demand, after a deliberate scope decision - not by default.
Development
npm install
npm run build
npm test
npm run lint # tsc --noEmit
Docker
The image listens on 0.0.0.0 inside the container (MCP_HTTP_HOST) and defaults to AUTH_MODE=gateway. Publish the port on loopback and pass a generated S2S secret. Startup rejects an empty CONDUIT_S2S_SECRET and known placeholders (your-s2s-secret, changeme, and similar), so each command below creates a value with openssl rand -hex 32 instead of embedding one. In gateway mode the AlertOps API key arrives on each request as X-AlertOps-Api-Key.
docker build -t alertops-mcp .
docker run \
-e CONDUIT_S2S_SECRET="$(openssl rand -hex 32)" \
-p 127.0.0.1:8080:8080 \
alertops-mcp
Compose refuses to start when CONDUIT_S2S_SECRET is unset. The Compose service is gateway mode, sets no ALERTOPS_API_KEY, and publishes 127.0.0.1:8080:8080. Export one secret when a gateway must sign requests with the same value the container receives:
export CONDUIT_S2S_SECRET="$(openssl rand -hex 32)"
docker compose up
Single-tenant HTTP (AUTH_MODE=env) still requires the S2S secret, then uses ALERTOPS_API_KEY from the environment. Export that key before running the command.
docker run \
-e CONDUIT_S2S_SECRET="$(openssl rand -hex 32)" \
-e AUTH_MODE=env \
-e ALERTOPS_API_KEY \
-p 127.0.0.1:8080:8080 \
alertops-mcp
Local HTTP without a secret is development-only. It binds 127.0.0.1 even if MCP_HTTP_HOST is 0.0.0.0, and logs a warning:
MCP_TRANSPORT=http MCP_ALLOW_INSECURE_DEV=1 npm run start:http
License
Apache-2.0
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.