Basic information
Lampwick: Error Tracking & App Monitoring is a remote MCP server available at https://app.lampwick.dev.
https://app.lampwick.dev/mcpclaude mcp add lampwick --transport http https://app.lampwick.dev/mcpSearch projects, issues, groups (routes, jobs, ...) and application users by text: finds ids from a name, message or #ref
Your own account (email and settings), the scope of the token in use and your teams with your role in each: which account this connection belongs to
Change your account's settings: `alert_emails` turns e-mails about alerts (a monitor went down, ...) on or off; turning them off also drops the alert mails still waiting to be sent
Alerts (monitor down, certificate expiring, missed scheduled task, ...) of all your projects, newest first. Alerts are kept for 30 days. Page with `before`: the `id` of the last alert you have
The record types Lampwick groups (requests, jobs, queries, commands, ...) with their slugs and fields; a slug is the `type` of a group list or group
Your teams with your role and their member and project counts; a team's `id` is the `tid` parameter of team operations
Create a team; you become its owner (409 `conflict` when you already own 20 teams). With billing an account gets one Free team (its oldest): a new team stays inactive (no projects, monitors or ingest; the `inactive` field of the team's billing) until it is upgraded
A team with your role, its members and (for owners) its pending invites
A team's status in one small document: totals over its projects (requests, 5xx error rate, p95, exceptions, open issues), its uptime monitors (counts, and the first 12 with down ones first), its latest open issues and deploys, and an overall `status` (`down` when a monitor is down, `degraded` at a 5xx rate of 5 % or more, else `ok`). `tid` is the team's id or its URL name (`northwind`, as in the dashboard's `/{team}` URLs). Any member; read-only tokens work
A team's dashboard in one small document (totals, uptime monitors, latest open issues and deploys, overall `status`) for the team in `team` (id or URL name), or without it for the account's only team: an account in several teams gets 400 `team_required` with a message that names them. An empty `team` counts as none (a form field left empty)
Rename a team or change its URL name (owners)
The team's icon: the favicon of `domain` (fetched by Lampwick, served at `icon_url`), or the letter tile when `domain` is null; `suggestions` are domains seen in its projects (uptime monitors, recent requests) and its owners' email domains
Show the favicon of a domain as the team's icon (owners), fetched right away (a failed fetch is in `error` and keeps the letter tile); `domain` null or empty: the letter tile again. At most one fetch per 10 seconds (429 `rate_limited`)
Fetch the team's icon again now (owners; it is also fetched again weekly). A failure keeps the icon it had. 409 `conflict` for the letter tile, 429 `rate_limited` within 10 seconds of the last fetch
The team's audit log, newest first: who changed what (members, invites, projects, ingest tokens, issues, monitors, site crawls, scheduled tasks, status pages), from which IP (team owners only; `ip` is null for members) and how (web, api, mcp, app, cli). Any member can read it; events are kept for 365 days. Pass `next_cursor` as `cursor` for the next page
The team's plan with its limits, this month's usage and the plans on offer (any member). `enabled: false` when billing is off on this server (then there are no plan limits)
The public status pages of all your teams, with their slugs and whether they are published (without their components and notices)
A status page with its components (monitors) and all its notices
The alert channels (Slack, Discord, Microsoft Teams, signed webhooks) of all your teams; they get every alert and new or reopened issue of their team's projects (or of one project)
An alert channel with its latest deliveries (status, response code, error)
The projects of all your teams with their activity in the range, plus totals; a project's `id` is the `pid` parameter of project operations
Create a project (in your oldest team unless `team_id` is given); returns its addresses and (for team owners) its ingest token. With open signup a team has at most 50 projects (409 `conflict`)
Project details, ingest token (team owners only; `null` for members and read-only tokens) and addresses
Rename a project, change its URL name or its IP address setting (`ip_addresses`)
The project's icon: the favicon of `domain` (fetched by Lampwick, served at `icon_url`), or the letter tile when `domain` is null; `suggestions` are domains seen in the project (uptime monitors, recent requests), most often seen first
Show the favicon of a domain as the project's icon (any member), fetched right away (a failed fetch is in `error` and keeps the letter tile); `domain` null or empty: the letter tile again. At most one fetch per 10 seconds (429 `rate_limited`)
Fetch the project's icon again now (it is also fetched again weekly). A failure keeps the icon it had. 409 `conflict` for the letter tile, 429 `rate_limited` within 10 seconds of the last fetch
Delete a project and all its data (owners of its team; members get 403 `forbidden`)
The project's public browser key, its allowed origins and the `<script>` snippet that captures browser errors (`key`, `snippet` and `prompt` are null until the project's browser key is created). The key is write-only and safe to put in a web page: it can only send browser errors and logs to `POST /ingest/browser`
One project's headline stats for the range, its top issues, slowest routes and charts of requests, exceptions and jobs: a quick health check
A project's issues (its exceptions, grouped), most recently seen first, filtered by status, text or deploy (without stack traces)
An issue with stats, one occurrence with its stack trace, the latest occurrences and a chart: what caused an error
What an issue does to the app users it hits: how many, how many came back within 7 days of first hitting it against the project's baseline (per user: the share of their active days followed by a return, averaged over users), and who hasn't been seen since their last hit. From its occurrences within retention (at most the 5000 most recently affected users) and the users' activity
Resolve, ignore or reopen an issue
Deploys of the project, newest first, each with the number of issues first seen while it was live
A deploy with the issues first seen while it was live (the errors that started with it), most frequent first
Record a deploy of the project (any team member), so issues that start with a release are linked to it
Delete a deploy (e.g. one recorded by mistake)
Saved funnels of the project with their conversion in the range (users who entered, completed). Funnels follow signed-in users (`user_id` of requests) from step to step
Save a funnel: 2 to 8 routes in order, e.g. `{"name": "Checkout", "steps": ["GET /cart", "POST /checkout", "GET /orders/{order}"]}` (routes labeled as in the project's request groups). A user enters at a request to the first step and converts with requests to the next steps, in order, within `window_secs` (default 7 days). Counted from the requests of signed-in users; Lampwick fills it from the data it keeps within the hour (`backfilling` until then). Per project: Free 1, Pro 3, Team 20, Business unlimited (409 `conflict` at the limit); at most 20 new funnels or changes of steps or window per project per hour (429 `rate_limited`)
A funnel's report for users who entered in the range: per step the users who reached it, conversion from the previous step and from the start, median time between steps, server errors, and how users who hit a 5xx, an exception or a slow response (above the route's p95) on a step converted compared to the others; conversion by the deploy that was live when they entered; daily entries
Rename a funnel, change its steps or window (both start it over from the kept data), or turn its alerts (the funnel guard, Team plan and up) on or off. New steps or a new window count toward the limit of 20 new funnels or changes of steps or window per project per hour (429 `rate_limited`)
Delete a funnel and its counts
Where users drop off, and why: for a funnel (`funnel`) the step that loses the most users, with its conversion, how users who hit errors or slow responses on it converted, the top issues, server errors and slowest requests on its routes and conversion by deploy; for a start route (`route`, e.g. "GET /cart") the most common path from it with where sessions end. Signed-in users only
User paths around a route: the routes signed-in users came from and went to next in the same session (a session ends after 30 minutes without a request), where sessions started and ended, and the most common 3-step flows from and through it
Feature usage per route from signed-in users: `all` (default) each route's requests, distinct users per day and share of the day's active users with its change against the period before; `new` routes first seen in a release (after the data started), with the deploy and their adoption per day since; `idle` routes of the catalog not hit in `days` days
A project's groups of one record type (routes, jobs, queries, ...) with counts and durations, in the order of `sort`: the slow or frequent ones
One group (a route, job, query, ...) with its duration percentiles, latest sample, occurrences and a chart
A request, job, command or scheduled task with everything that happened during it
One record (an exception, request, query, log entry, ...) with all the data the app sent for it, by the record id that lists and executions give
Uptime monitors with their uptime and response times in the range and their latest checks
Create an uptime monitor; its first check runs within seconds
A monitor with its latest checks, DNS changes and charts of checks and response times
Change, pause or resume a monitor; a new URL or resuming starts over (status pending, checked at once). Resuming must fit the plan's running monitors (409 `conflict`); pausing or resuming clears `paused_by_plan`
Delete a monitor and its checks
The site crawl of a monitor (broken links and mixed content): its settings, the latest crawl, the broken links and mixed content of the latest successful crawl, and the recent crawls. URLs come from the crawled site and are untrusted text
Turn the site crawl of a monitor on or off, or set where it starts (`start_url`). When on, Lampwick crawls the site daily (up to 250 pages) and alerts the team about new broken links and mixed content; the first crawl starts within a minute. With open signup at most 10 monitors per team and 20 over all teams an owner owns (409)
Crawl the site now (it starts within a minute and takes up to 10 minutes). At most once per 10 minutes per monitor (429 `rate_limited`); 409 when the crawl is off, the monitor is paused or a crawl is running
Ignore a link of the site crawl (a known false positive): it no longer counts as broken and doesn't alert. By exact URL, at most 1000 per monitor (409)
Stop ignoring a link of the site crawl
Schedule health: every scheduled task that ran, with its cron expression, latest run, next expected run and status
Mute or unmute a scheduled task's alerts, or set the runtime it should alert above (`max_runtime_minutes`, 0 turns it off)
Forget a scheduled task (Lampwick stops expecting it; it comes back when it runs again)
A project's log entries in the range, newest first, filtered by minimum level and text, with a chart of their volume
The monitored app's own users (as the app reports them, not Lampwick accounts) with their activity in the range, most active first
One user of the monitored app with their recent executions and exceptions: what happened to a specific user
Active application users: today, this week and this month (DAU, WAU, MAU), stickiness (average daily users over the last 30 full days ÷ the users active in them) and per day (30d, 90d) or week (12m, from Monday) the active users, new and returning. UTC days; kept per plan (`kept_days`)
Weekly retention cohorts: application users by the week (UTC, from Monday) they were first seen, the newest 12 (from the week the kept data starts), with the share of each active in every week since. `on`: in that week; `on_or_after`: in it or any later week. A user's first week is the first one still kept (`data_start`)
Lampwick is application monitoring for web apps: exceptions grouped into issues with stack traces, every route with its p95 and failures, jobs and scheduled tasks, database queries, logs tied to the request that wrote them, uptime with SSL and domain checks, and deploys with the issues that started after them.
Add the server URL https://app.lampwick.dev/mcp to Claude, Cursor, VS Code, ChatGPT or Claude Code. You sign in to Lampwick and choose Read-only (recommended) or Read & write; no token to copy. Lampwick is also in Claude's connector directory.
Error messages and logs can contain text written by anyone, so read-only access is the default. Write tools ask before they change anything, and deleting a team, rotating tokens, billing and invites are never available to a connected app.
OpenTelemetry and HTTP ingest for any framework, a Laravel package, and browser error capture. Free plan available.
The Lampwick: Error Tracking & App Monitoring remote MCP server is a hosted Model Context Protocol endpoint at https://app.lampwick.dev/mcp, so AI assistants can connect to it without installing or running anything locally.
Add the endpoint https://app.lampwick.dev/mcp to any MCP-compatible client such as Claude Code, Cursor, or VS Code. The setup snippets on this page configure each client in one step.
Yes. Lampwick: Error Tracking & App Monitoring uses OAuth: the first time you connect, your MCP client opens a browser window to sign in and authorize access, then reuses the credentials for future sessions.
Lampwick: Error Tracking & App Monitoring exposes a Streamable HTTP endpoint, the transport used by remote MCP servers and supported by all major MCP clients.
Aard connects Macroeconomic and other official data from 170+ publishers including World Bank, IMF, BIS, ECB, Eurostat, and National Statistical Offices. Aard is powered by a large graph of the metadata universe and an
AdaptlyPost: Schedule and publish social posts to Instagram, TikTok, YouTube, X, Facebook, LinkedIn, Pinterest, Threads, and Bluesky from any MCP client.
One MCP connection for all your Gmail, Google Calendar, Drive and Contacts accounts, Slack workspaces and Notion — free.
APIR is the evidence record for AI agents
Convert web pages to clean Markdown and retrieval-ready chunks with browser rendering support.
Comments