Error tracking for web apps and APIs

Lampwick records every exception in your app (Laravel with one package, any other stack over HTTP) and groups them into issues by class, code, file and line, so a thousand identical errors are one row with a count. Each issue links to the request or job it happened in, with the queries, logs and outgoing calls around it. The Laravel package also records your requests, queries, jobs, scheduled tasks and logs, and an AI agent can read all of it over the MCP server and propose a fix.

What you get

• Issues. Exceptions grouped by class, code, file and line, with the stack trace and source code, the users affected, an Unhandled badge, and the context of an occurrence: Server, Deploy, User and During. A resolved issue reopens when it comes back, and the team gets a push about the regression.
• Requests. Every route with its status codes, p95 and failures, and a timeline of what each request did.
• Queries. Database queries grouped, with average and p95 duration, next to outgoing HTTP, cache, mail and notifications.
• Jobs. Every queued job and command: attempts, failures and duration, with the exception that stopped it.
• Logs. Every log line tied to the request or job that wrote it, searchable by level and text.
• Deploys. Each deploy, with the issues first seen after it.
• Browser errors. One script tag turns JavaScript errors in your visitors' browsers into issues too (browser errors).
• Notifications for new and returning issues by push in the iOS app and to Slack, Discord, Microsoft Teams or a webhook (alert channels).

You can see the real app with example data in the demo.

Your agent reads the errors

Connect Claude Code, Cursor, Codex or another MCP client to https://app.lampwick.dev/mcp, read-only by default. Then ask from your repository:

Which open issues in this project were first seen since 08:00 UTC today?

Show me the stack trace of #209 and the request it happened in. Propose a fix and a test. Don't commit.

The agent reads the issue, opens the file and line it points to, and proposes a change you review. Setup per client: connect Lampwick to your AI agent. A full walkthrough: debug Laravel production errors with Claude Code.

Set it up

composer require lampwick/laravel
echo LAMPWICK_TOKEN=YOUR_PROJECT_TOKEN >> .env
php artisan lampwick:agent

YOUR_PROJECT_TOKEN is on the project's Settings page. The app hands records to the ingest daemon on 127.0.0.1:2407, so requests never wait on the network. Run php artisan lampwick:test to check the connection, and keep the daemon running under Forge, Supervisor or systemd: connect a Laravel app.

Not on Laravel? Any PHP app, other language or script can send exceptions and logs as JSON over HTTP ingest.

What it doesn't do

• The package supports Laravel 12 and 13 on PHP 8.2 or later. Other PHP frameworks use HTTP ingest, which means writing the records yourself.
• Browser stack traces point into the bundled code: there are no source maps.
• No distributed tracing across services, no profiling, no session replay.
• New issues notify by push and alert channels, not by e-mail.
• History follows your plan: 3 days on Free, 14 on Pro, 30 on Team, 60 on Business.

If you compare with other tools, read Lampwick vs Sentry and Lampwick vs Laravel Nightwatch.

What it costs

Priced per team by events: every exception, request, query, job and log line is one.

Uptime monitors are included on top. Browser errors have their own budget of 10% of the events. See pricing.

Related

• Common Laravel and PHP errors, with causes and fixes: errors, such as 419 Page Expired and Allowed memory size exhausted.
• What to log and where: Laravel logging in production.
• Failing queued jobs: why Laravel queued jobs fail.
• The scheduler: cron job monitoring.

Start free and connect your Laravel app.