New advertisers: 50% off your first month with code FIRSTMONTHSee slots

Daemon-based GHCi build status monitor for Haskell projects, using an MCP server

GitHub stars
71
View on GitHub

Add marketplace

/plugin marketplace add atelier-hub/tricorder

Run it inside Claude Code once, then install the plugin below.

Install

/plugin install tricorder-mcp@tricorder

Run it inside Claude Code after adding the marketplace.

Files

Discovered from source

Using the tricorder MCP Server

Use the mcp__plugin_tricorder-mcp_tricorder__* MCP tools for GHCi build status, diagnostics, and test results in this project — don't run tricorder as a Bash command; the MCP server already wraps it.

Tools

  • start — start the daemon (no-op if already running)
  • stop / restart — stop/restart the daemon; pass force: true to skip waiting on pending queries
  • status — current build status: diagnostics, errors, warnings. Pass wait: true to block until an in-progress build finishes, verbose: true for full GHC message bodies, or expand: N to expand a single diagnostic. Returns JSON.
  • test_results — output from the latest test run. Pass wait: true to block until the build/test cycle finishes, or failed: true to show only failing suites.
  • source — Haskell source for one or more project or installed modules (e.g. Data.Map.Strict or Data.Map.Strict#insert). Prefer this over downloading tarballs.
  • eval_comments — eval comments and their evaluated results from the latest build. Pass wait: true to block until the build finishes. Returns JSON.
  • log_contents / log_path — the daemon's log output, or the path to its log file.

All tools accept an optional projectRoot (defaults to the current working directory).

Workflow

  1. Edit source files.
  2. Call status with wait: true — blocks until tricorder finishes recompiling.
  3. If errors are shown, fix them and repeat.
  4. A clean result means the build is clean; configured test suites run automatically.
  5. Call test_results with wait: true, failed: true to see any test failures.

Notes

  • The daemon is per-project, scoped by the resolved project directory.
  • status auto-starts the daemon if it isn't running.
  • Do not run cabal test manually — the daemon manages test execution after each clean build.
  • If these MCP tools are unavailable, disconnected, or misbehave, and the tricorder CLI is installed, fall back to running it directly via Bash (see the tricorder skill from the tricorder plugin, if installed).

From the atelier-hub/tricorder README

Tricorder

Tricorder aims to empower users developing programs with Haskell and LLM coding agents. It does so by providing operations to surface the right information required at a given stage: documentation, build status, diagnostics, etc.

ℹ️ More information about how to use tricorder can be found in this presentation for the Haskell Foundation's "Haskell and AI Workshop".

Like similar tools (ghcid, ghciwatch), it builds the code continuously on every change, presents diagnostics, and runs the tests afterwards. However, Tricorder offers other advantages:

  • Designed for humans - A tricorder ui interactive TUI mode that presents stats in real time for developers.
  • Designed for agents - A SKILL is provided to inform agentic usage via the tricorder CLI.
  • Background builds - Building in the background using a daemon allows different clients to query the build state simultaneously without triggering multiple rebuilds. For instance, we ship the tricorder ui TUI and the tricorder status CLI command that communicate with a single daemon via a socket.
  • Sane defaults - Running tricorder start should Just Work™ for most cabal- and stack-based Haskell projects.
    • Daemon restarts automatically when cabal files change
    • If customization is needed it can be provided at different levels via a .tricorder.yaml or CLI args.
      • Optional config includes which cabal packages to watch, which exact command to use to enter a GHCi session, customizable key bindings, etc.
  • Multi-package projects - For projects containing multiple packages, either described through cabal.project or stack.yaml, Tricorder discovers every package's components automatically, building and running tests across all of them. No manual target configuration needed!
  • Project context - Tools like tricorder source Some.Module will attempt to find and provide the source code for a given dependency from disk, which allows exploring library APIs more easily.
  • Machine-readable output - Using tricorder status --json we can get build information in a format appropriate for programmatic usage.

For a deeper guide on the various features of Tricorder, see Features of Tricorder.

Quick start

If you have [Nix] installed, you can quickly test Tricorder out with the following command:

nix run --accept-flake-config github:tweag/tricorder -- ui

--accept-flake-config makes Nix use the binary caches declared in the flake (tweag-tricorder.cachix.org and cache.iog.io). Without it, Nix builds Tricorder and the Haskell toolchain from source.

Usage

With the tricorder CLI in your path, start the Tricorder daemon either explicitly with tricorder daemon start or automatically by just running tricorder ui, which will start the daemon if it's not already running and show you the TUI.

tricorder --help will show you a detailed description of each command available.

For using Tricorder with Claude Code, Copilot, or another coding agent, see Using with coding agents.

Installation

Tricorder is supported on both Linux and macOS.

Tricorder is published on both on Hackage and on Stackage from Stackage Nightly 2026-08-19 and onwards, so you may install it using either cabal or stack, whichever is your preference:

cabal install tricorder
# Or
stack install tricorder

On NixOS, when installing with cabal, you might get errors like this:

Missing dependency on a foreign library:
* Missing (or bad) C library: z

Pass the following flag to ensure zlib uses the zlib bundled with your system: --constraint='zlib +bundled-c-zlib'.

To install and use Tricorder in your NixOS configuration, see Using with Nix.

Tricorder is also released as pre-built binaries on GitHub.

Configuring

See Configuring Tricorder

Development

nix develop
tricorder ui

Libraries

This repository also contains Atelier, a set of Haskell libraries providing foundational infrastructure for effect-based applications (to be extracted into their own repository).

Some less maintained Atelier libraries have been moved to atelier-extras.

Details

Marketplace
atelier-hub/tricorder
Source
agent-plugins/tricorder-mcp
Category
Community

More from atelier-hub/tricorder

All 2

Similar to tricorder-mcp

More MCP Servers