The Model Context Protocol (MCP) is an open standard for connecting AI assistants to external tools and live data. Statusfield hosts a read-only MCP server for catalog search and recorded status from services your workspace monitors. Ask a compatible assistant questions like "Which of my monitored services are reporting problems?" The tools read existing records; they do not run a new probe or identify the cause of an application failure.
Requirements
The MCP server uses the same sf_ API key as the REST API — no separate credential. Create one at Integrations → API Keys, where you can also revoke it later. The key is shown only once at creation time.
Starter plan and up
MCP access is available on the Starter, Pro, and Team plans — the same
gate as the REST API. It is not available on the Basic plan. A Basic-plan
key returns 403 plan_restriction. See
/pricing for per-plan limits.
Connecting
The server speaks Streamable HTTP at POST /api/mcp and authenticates with your sf_ key in the Authorization header.
Get an API key
Go to Integrations → API Keys at statusfield.com/integrations and create a key. Copy it — it is shown only once.
Add it to the Claude Code CLI
Run the following, replacing sf_<your-key> with your key:
claude mcp add --transport http statusfield \
https://statusfield.com/api/mcp \
--header "Authorization: Bearer sf_<your-key>"Or connect another MCP client
Use your client's supported remote Streamable HTTP connection with URL
https://statusfield.com/api/mcp and header Authorization: Bearer sf_<your-key>.
Client setup varies; use its documented remote-server settings. Do not paste
your key into a prompt or a shared configuration.
Ask a question
Once connected, ask your assistant something like "Is GitHub having any issues right now?" for a service your workspace monitors. The assistant can use the matching tool; review the result before acting on it.
Tools
The server exposes three read-only tools:
| Tool | Input | Description |
|---|---|---|
search_services | query: string | Search the catalog by name; returns up to 10 matches as Name (slug) lines |
get_service_status | slug: string | Recorded status of a monitored service and its components |
list_my_down_services | (none) | Your own monitored services currently down or degraded |
Catalog search vs. monitored status
search_services searches the public catalog without exposing status.
get_service_status returns status only for services your workspace actively
monitors. list_my_down_services reports your own monitors that are down or
degraded. For the full monitor list, use GET /api/v2/monitors.
An empty problem list means no matching records were returned. It does not establish that every dependency or user workflow is healthy. MCP status text does not include observation timestamps; use the REST API or service detail page when you need to check freshness.
Rate limits & quota
- Rate limit: your plan's per-workspace burst limit applies to every request.
X-RateLimit-LimitandX-RateLimit-Remainingare returned on responses, and aRetry-After(seconds) header is added on429. - Monthly quota: only
tools/callinvocations count toward your plan's monthly request quota. Protocol traffic (initialize,tools/list) is free.X-Quota-*headers are returned on metered requests, and exceeding the cap returns429 quota_exceeded. The counter resets at the start of each UTC month.
Current limits, shared with REST per workspace:
| Plan | Requests / minute | Requests / second | Requests / UTC month |
|---|---|---|---|
| Basic | — | — | No API access |
| Starter | 30 | 5 | 100,000 |
| Pro | 60 | 10 | 500,000 |
| Team | 120 | 20 | 2,000,000 |
Example prompts
Example questions and the tools an assistant can use:
- "Is GitHub having any issues right now?" →
get_service_statuswith sluggithub - "Is anything I monitor down right now?" →
list_my_down_services - "Find monitoring services" →
search_serviceswith querymonitoring
Next steps
- API reference — the full REST API, including
GET /api/v2/monitorsfor your own workspace's monitors. - Pricing — per-plan rate limits and monthly quotas.