What the server exposes

The Loggerhead MCP server is a small program, loggerhead-mcp. It speaks the Model Context Protocol over stdio to your coding tool, and it reads data from the Loggerhead engine over the local HTTP API. It never opens the database file itself.

Tools

Tool	Purpose
query_logs	Search and filter log records by text, severity, and time.
query_metrics	Read time-series metrics by name pattern and time.
query_traces	Search traces by service, operation, duration, and time.
get_trace	Read every span of one trace by trace ID.
get_receiver_info	Report the effective OTLP ports of the root receiver, the conflict flags, and the addresses to send to.
query_improvements	List the telemetry-quality findings from the Linting page, with filters for impact, source, and kind. A folded group carries its members. The result also lists the findings that resolved in the last 24 hours, with the reason.
list_environments	List every workspace with its environments, their deployment.environment.name tags, the receiver ports per workspace, and the pair the app reads now.
list_lint_rules	List the lint rules behind the Linting page: each rule’s state, its thresholds with the value in force and the default, its exceptions, the ignored findings, and the recent changes with their source and reason.
query_ai_findings	List the findings of the last AI analysis run on the AI analysis page, with the status, the time, the window, and the model of the run. It does not start a run.
set_lint_rule_enabled	Switch a lint rule on or off for a workspace.
set_lint_threshold	Set one threshold of a lint rule.
reset_lint_threshold	Put one threshold back to its default.
add_lint_exception	Hide the findings of one rule that match a service, a metric, an attribute key, an operation, or a log message.
remove_lint_exception	Remove one exception.
ignore_finding	Ignore one finding by its id.
unignore_finding	Stop the ignore of one finding.
mark_finding_fixed	Mark one finding fixed after a code change, so Loggerhead judges it by the telemetry after the fix.

The seven tools that change the lint rules need a reason. The rule settings belong to the workspace, so every environment of it uses a change. The Rules tab of the Linting page shows each change with its source (MCP or the app), its time, and its reason, and you can undo it there.

Workspaces and environments

The app reads one environment at a time, and so does every tool. Each tool takes two optional arguments, workspace and environment. Omitted, they mean the workspace and the environment the app shows. Name a workspace by its name or its slug, and an environment by its name, its slug, or its deployment.environment.name tag. Call list_environments first when you need the names. See Workspaces and environments.

Resource

Resource	Content
loggerhead://stats	Total counts of logs, metrics, and spans in the database.

What it cannot do

• It cannot delete data or send telemetry. The only settings it changes are the lint rules of a workspace, and each change has a reason that you can see and undo on the Rules tab of the Linting page.
• It cannot start an AI analysis. A run can send telemetry to a cloud model, so only the user starts it, in the app. query_ai_findings reads the last run.
• It cannot reach another machine. It talks to 127.0.0.1 only.
• It cannot see data that is not in Loggerhead. If a service does not send to the root, the tool cannot find it.

Why it goes through the engine

DuckDB allows one process to hold the file. The engine holds it. The MCP server, the app, and any script all read through the engine’s API. That is what lets the app and your coding tool work at the same time.