@brass-lantern/mcp
v2.3.1
Published
A Model Context Protocol server for brass-lantern: lets a coding agent audit, play and read the docs for the text-adventure world it is writing.
Maintainers
Readme
Brass Lantern MCP server
@brass-lantern/mcp: a local Model Context Protocol server that lets a coding agent write, audit and play a Brass Lantern world, and read the docs for the version it is using.
Install
In Claude Code:
claude mcp add brass-lantern -- npx -y @brass-lantern/mcpOr in any client’s JSON config:
{ "mcpServers": { "brass-lantern": { "command": "npx", "args": ["-y", "@brass-lantern/mcp"] } } }Node 24 or later. World paths are relative to where the client starts the server. Worlds must be inside that directory or one you allow with --root <dir> (repeatable). Each audit_world and play call runs in its own worker thread and is stopped after 10 seconds; --timeout <seconds> changes that.
Tools
audit_world: checks a world file for broken exits, bad conditions, missing references and more.play: types commands into a fresh, seeded game and returns the transcript and final state.docs: the documentation for this version, as an index, a page, or a search.
It uses the world’s own project’s engine when it has one, and every reply says which engine ran.
Safety
It runs the world module it is pointed at, with your permissions, like your test runner. It is local stdio only and makes no network calls. The allowed-directories check confines which world file is loaded, not what that module imports or does: any module inside an allowed directory can be run. It is not a sandbox.
Docs
Coding agents covers the tools, the engine choice and the limits in full. Source and issues: github.com/mrballistic/brass-lantern. MIT licensed.
