Delv
CommunityAbandoned· 11mo4.3by sonirico

MCP Stockfish

Bridges AI systems to the Stockfish chess engine with concurrent sessions and full UCI command support.

C
Safety & Trust

Delv Safety Grade: C

Score 58/100 · assessed 2026-04-28

Maintainer45
Permissions65
Supply chain40
Transparency70
Incidents100

MCP Stockfish is a community-maintained bridge connecting AI systems to the Stockfish chess engine via UCI protocol. The maintainer 'sonirico' appears to be a solo developer with limited GitHub presence and sparse activity across repositories. Installation requires 'make install' which typically compiles from source, introducing supply chain concerns as there is no package registry distribution or dependency pinning. The server spawns local Stockfish processes and manages concurrent chess sessions, requiring shell execution and filesystem access to the engine binary. Permissions are moderately scoped to chess engine interaction rather than broader system access. The repository is open source with reasonable documentation of UCI commands and session management. No security incidents are known. The narrow use case (chess analysis) limits blast radius, but the custom build process and solo maintainer present moderate supply chain and maintenance risks.

Lethal Trifecta (prompt-injection exposure)

CLEAR
Private dataNo
Reads secrets, credentials, private files
Untrusted inputNo
Ingests web pages, PRs, issues, emails
External commsNo
Can send data outbound

Pure chess engine. No I/O.

Green flags

  • Narrow scope limited to chess engine interaction reduces attack surface
  • Open source with clear UCI protocol documentation
  • No known security incidents or malicious behaviour
  • Concurrent session management suggests thoughtful architecture

Red flags

  • Solo maintainer with limited GitHub activity and unclear bus factor
  • Custom make-based install with no package registry distribution
  • Spawns local processes (Stockfish binary) requiring shell execution
  • No dependency pinning or reproducible builds visible
  • Limited community adoption and review for supply chain validation

Permissions requested

Shell executeRead filesOutbound network
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Install

make install

Review

This is a Model Context Protocol server that lets Claude talk directly to Stockfish, the open-source chess engine that outplays every human on the planet. It handles concurrent sessions, so you can analyse multiple positions at once without stepping on your own toes, and exposes the full UCI command set. That means you get best move suggestions, position evaluations, and the ability to run endgame studies or prep for tournaments without leaving your chat window. I'd reach for this when I'm annotating games or testing opening ideas. You can feed Claude a PGN, ask it to critique your moves, and it'll query Stockfish for the engine's take on each position. The concurrent sessions matter if you're comparing lines side by side or running a batch of tactical puzzles. The UCI support is complete, so you can set search depth, tweak contempt values, or enable MultiPV to see the top three candidate moves. It's not a GUI replacement, but it's faster than alt-tabbing to Lichess or Arena when you're already working in Claude. Quirks: you need a local Stockfish binary installed and in your PATH. The README assumes you know what UCI is and how chess engines work. If you've never run `stockfish` in a terminal, you'll want to skim the Universal Chess Interface spec first. The install is a Makefile, which is fine on macOS or Linux but might trip up Windows users who aren't running WSL. No environment variables to configure, which is refreshing, but you do need to point the config at the right executable path if Stockfish isn't already discoverable. This isn't for casual players who just want to know if their last move was a blunder. It's for people who already use engines and want to script analysis workflows or integrate chess into a larger AI pipeline. If you're preparing for a tournament, running post-game reviews, or building a chess tutor bot, this is a clean way to wire Stockfish into Claude. If you just want to play a game, open Lichess.
Verdict

Install this if you're a serious chess player or coach who already uses Stockfish and wants to automate analysis inside Claude. Skip it if you don't know what UCI commands are or if you're looking for a beginner-friendly chess tutor. It's a bridge for people who know both ends of the span.

Good at

  • Full UCI command support means you can tweak search depth, MultiPV, and other engine parameters without hacks.
  • Concurrent sessions let you compare multiple positions or lines in parallel without restarting the engine.
  • No environment variables to configure, just point at a Stockfish binary and go.
  • Clean separation of concerns: it's a protocol bridge, not a chess GUI pretending to be something else.

Watch out

  • Requires a local Stockfish install, which adds a setup step and assumes you know how to manage binaries.
  • The Makefile install might confuse Windows users who aren't running WSL or a Unix-like shell.
  • No built-in rate limiting or depth caps, so a careless query can peg your CPU for minutes.
  • Documentation assumes familiarity with UCI and chess engine workflows, not beginner-friendly.

Use cases

  • best move analysis
  • position evaluation
  • endgame studies
  • tournament preparation

Getting started

1. Install Stockfish on your system and confirm it's in your PATH by running `stockfish` in a terminal. 2. Clone the repo and run `make install` to build the MCP server. 3. Add the server to your Claude Desktop config, pointing the command at the built executable and ensuring Stockfish is discoverable. 4. Open Claude Desktop and ask it to evaluate a chess position in FEN notation to verify the connection works. 5. Watch out for the search depth: Stockfish can chew CPU if you ask for deep analysis without setting a limit.

Works with

Claude DesktopClaude CodeCursor

Similar MCPs