Hone

Hone is self-hosted performance telemetry for Laravel apps. Your apps send the telemetry they already produce through Laravel Nightwatch, including requests, queries, jobs, outgoing HTTP calls, exceptions, cache and log activity, to a Hone deployment you run. Hone stores it in that deployment's PostgreSQL database and rolls it up into daily aggregates. You read it back through an MCP server, so a coding agent such as Claude Code can answer questions like "what should we improve this week?" or "what got slower after the last deploy?"

Hone is open source under the MIT license. It is built on laravel/nightwatch: a small client package replaces Nightwatch's transport so records go to your Hone server instead of Nightwatch's hosted service. You do not need a Nightwatch subscription, and Hone never contacts Nightwatch's cloud.

What Hone does not do

Hone is deliberately narrow. It has no web UI or dashboard, no alerting or paging, no distributed traces, and no raw request bodies. Individual events are kept for 72 hours by default; after that only the daily aggregates remain. It only monitors Laravel apps. Sampling and redaction stay in each app's Nightwatch configuration.

One Hone deployment serves one team. Any number of that team's apps can report to it, and anyone holding an MCP credential for it can read all of them.

How it fits with Scalpels

Hone is a provider app. Scalpels installs Hone in your own GitHub and Laravel Cloud accounts, then attaches your other Laravel apps to it: it mints a per-app ingest credential on Hone, writes the connection values into the app's environment, and opens a draft pull request that adds the client packages. Your agent can also reach Hone's MCP tools through the Scalpels MCP server, without a separate Hone credential.

In these docs

• Getting started: install Hone through Scalpels and check that it is running.
• Connecting your apps: attach a Laravel app so it reports to Hone, and prove the connection works.
• Connecting your agent: reach Hone's MCP server through Scalpels or with a direct MCP credential.
• MCP tools: the 24 read-only tools, their arguments, and how to read their answers.
• Data and retention: how telemetry is stored, rolled up, compared across deploys and pruned.
• Configuration: the optional settings on the Hone server and in a connected app.
• Troubleshooting: the failures you are likely to see and how to fix them.

Want a branded address for this deployment? Use your own domain.