MCP · production-ready

MCP server for event logs

Connect Claude or ChatGPT to production events with one npx command. 18 tools for search and AI-assisted analysis.

Free: 24h of event retention Documentation

SDKs for popular stacksOIDC OAuth 2.1Managed Cloud

Quickstart

Connect in under a minute

Pick the most convenient method — CLI command, JSON config or Custom Connector.

Get an API key

Create a new key on the API keys page of your dashboard.

Open LK · API keys

Connect MCP

Pick a client on the left — we'll walk you through install and a live tool call on the right.

~/Library/Application Support/Claude/claude_desktop_config.json

Saved

Сохрани файл и перезапусти Claude Desktop — сервер появится в Settings → Developer.

Tool catalog

18 read-only tools

The AI agent sees each tool as a separate function: you ask in plain English, the assistant picks the right call and passes the arguments. No write operations — events and settings stay untouched.

Search and event access

· 5

The basics: find events by text or JQL, open a specific event, list apps, look up similar cases.

search_logs

Full-text event search with filters by level, time range, tags.

query_logs_jql

JQL expressions for precise queries: AND/OR, ranges, field projections.

get_log_details

Full event card: stack trace, context, environment, metadata.

list_apps

All apps available to the current API key — slug, name, status.

find_similar_errors

Events with a fingerprint similar to a given one — for finding repeats.

Analytics and trends

· 5

Time-series distributions, heatmaps by hour and weekday, breakdowns by any field, top messages and overall traffic.

error_histogram

Time-series error distribution with a configurable step (minutes/hours/days).

error_heatmap

Heatmap by hour × weekday — to spot load patterns.

error_stats_breakdown

Event-count breakdown by any field — browser, country, version.

traffic_stats

Overall event volume for a period: total, by level, by app.

top_error_messages

Top-N most frequent error messages with counts and share.

Grouping and correlations

· 3

Fingerprint clustering of recurring errors, correlations between event types, rebuilding the chain inside a single user session.

grouped_errors

Clusters of events with the same fingerprint — counts and first/last seen.

find_correlated_errors

Which other errors tend to appear together with the given one — by time windows.

errors_in_session

All events within a single user session — for replaying the scenario.

AI analysis and triage

· 3

Root-cause summary for a single event, auto-triage of critical incidents within a window, user-impact estimate.

explain_error

AI summary for an event: likely cause, context, recommendations — in Russian or English.

triage_recent_critical

Auto-triage of critical events within a window: what to look at first.

impact_analysis

How many users and sessions an error affected; impact estimate.

Regressions and comparisons

· 2

Compare the current window with a baseline, hunt regressions between releases — which errors grew after the deploy.

baseline_compare

Compare the current window with a baseline — what changed in volume and shape.

version_regression

Errors that grew after a new release — a diff between versions.