Tools reference

The MCP server exposes 6 tools. Each is a single round-trip to the Loguro logs API — no SQL, no streaming, no internal chaining. Claude composes multi-step investigations by calling them in sequence.

When LOGURO_PROJECT is set, the project parameter becomes optional and defaults to that slug.

query_logs

Query logs from a Loguro project with filters. Use this to search for errors, slow requests, or any log entries.

Example. User asks: “Show me errors from the payments service in the last hour where the user was on staging.”

{
  "level": ["error"],
  "from": "2026-05-09T11:00:00Z",
  "context": [
    { "key": "service", "value": "payments" },
    { "key": "env", "value": "staging" }
  ]
}

get_log_timeline

Get logs surrounding a specific log ID within a time window. Useful for understanding context around an error.

Example. User: “What was happening around log 01HX9Q...?”

{ "logId": "01HX9Q...", "window": 60 }

get_distinct_values

Get all unique values for a field under current filters. Returns { values: [...] }.

Example. User: “What environments do we ship logs from?”

{ "field": "context.env" }

group_logs

Group logs by a field. Returns { grouped: { <value>: [logs] } } for each requested groupValue.

groupValues is required — the API does not auto-discover groups. Use get_distinct_values first if you don’t know the values.

Example. User: “Bucket the last hour’s logs by level for error and warning.”

{
  "groupBy": "level",
  "groupValues": ["error", "warning"],
  "from": "2026-05-09T11:00:00Z"
}

get_slow_logs

Shortcut to find slow requests above a duration threshold.

Example. User: “Anything slower than 2 seconds today?”

{ "thresholdMs": 2000 }

sample_logs

Get a random sample of logs, optionally filtered. Good for getting a feel for what’s in a project.

Example. User: “Show me 20 random debug logs to see what’s flowing through.”

{ "n": 20, "level": ["debug"] }

Context filters

Several tools (query_logs, group_logs, sample_logs) accept a context parameter — an array of filter objects that target arbitrary context.<key> columns.

[
  { "key": "user_id", "value": "42" },
  { "key": "duration", "value": 500, "op": ">=" },
  { "key": "env", "value": "staging", "negate": true }
]

The default operator is =. negate: true flips the filter to an exclusion (under the hood it maps to notContext_<key>). When both op and negate are set, the negation takes precedence and the operator is dropped.