MCP
Foam exposes an MCP server so AI coding agents can query your observability data — logs, traces, metrics, and issues (exceptions) — directly.
search_tracesFind distributed traces by service, status, duration, or attributessearch_logsQuery structured logs with filters across all servicessearch_metricsRetrieve time-series metrics with aggregationsget_issueFetch a Foam issue and its root cause analysislist_servicesDiscover all instrumented services and their metadatacount_eventsCount logs, traces, or spans matching a querylist_attribute_keysExplore available attributes for filteringlist_metric_namesDiscover available metric namesCursor
Add to .cursor/mcp.json in your project (or global settings):
{
"mcpServers": {
"foam": {
"type": "http",
"url": "https://sdk.api.foam.ai/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN"
}
}
}
}
Claude Code
claude mcp add foam \
--transport http \
https://sdk.api.foam.ai/mcp \
-- --header "Authorization: Bearer YOUR_TOKEN"
Codex
export FOAM_TOKEN="YOUR_TOKEN"
codex mcp add foam \
--url https://sdk.api.foam.ai/mcp \
--bearer-token-env-var FOAM_TOKEN
Other
URL: https://sdk.api.foam.ai/mcp
Transport: Streamable HTTP
Header: Authorization: Bearer YOUR_TOKEN
Example usage
Agent
What's the trace for request abc-1234?
Foam
gateway → auth (3ms) → checkout (12ms) → payments (4,201ms timeout)
Agent
Show me card-store logs between 2:20pm and 2:25pm
Foam
[14:23:01] charge_id=cx_8f2k · duplicate=true · skip_write
Foam
[14:23:04] charge_id=cx_1a9j · pool_exhausted · retry_queued
Agent
What metrics spiked in the last hour for payments?
Foam
http.server.duration p99: 1.2s → 8.4s (+600%)
Foam
db.pool.active: 48/50 (96% saturated)