Serveur MCP pi-code

pi-code

Claude Code experience for the pi coding agent, in one package. Point pi at a project that already has a .claude/ directory and it reads your existing config: rules, commands, skills, hooks, output styles, MCP servers, and agents. It also adds the Claude Code features pi lacks: a todo overlay, checkpoints, memory, web search, subagents, and goals.

What a repository ships is treated as untrusted until you approve it: project MCP servers, hooks, agents, rules, output styles, commands and skills load only once you say yes. A headless run (pi -p) cannot ask, so an undecided project loads none of them there, which is stricter than Claude, where a headless run uses them without showing the dialog.

Requirements

pi >=0.80.5 (0.99.x recommended) and Node >=22.19 for current pi. On an older pi, pi-code warns once at session start.

Install

pi install npm:pi-code       # from npm
pi install -l npm:pi-code    # project-local instead, writes .pi/settings.json

Other sources:

pi install git:github.com/ilovepixelart/pi-code
pi install ./pi-code         # local checkout, then /reload after edits

One pi install and everything below loads on the next start. pi list shows what is installed, pi config toggles individual resources, and pi update pi-code upgrades it. Each feature is an extension under extensions/.

What it does

Each topic links to its own doc with the full contract and any divergences from Claude Code.

• Hooks — your .claude/settings.json hooks on every lifecycle event, with Claude's tool vocabulary, decision fields, and background hooks.
• MCP servers — every Claude config scope, all four transports, OAuth, managed policy, timeouts, prompts, and resources. On pi 0.99 and later the servers are handed to pi's own MCP support.
• Custom slash commands — .claude/commands with arguments, bash spans, @file inlining, frontmatter, and model invocation.
• Skills — .claude/skills discovery plus the same dynamic content commands get.
• Subagents — built-in and custom agents, background runs, per-agent memory, worktree isolation.
• CLAUDE.md, @imports, and rules — the context files and path-scoped rules pi does not load natively.
• Settings env and flags — env blocks from every settings scope, exported with Claude's precedence, plus Claude's --settings and --setting-sources flags.
• Output styles — replace semantics, bundled built-ins, /output-style.
• Persistent memory — a per-repository store with a session-injected index.
• Statusline — your Claude statusLine command with the documented stdin JSON.
• WebSearch / WebFetch — key-free search and SSRF-guarded fetch.
• Claude plugins — installed marketplace plugins: commands, agents, hooks, MCP servers, styles, skills.
• Goal: /goal <condition> keeps the session working until a separate model check confirms the condition holds, with status, clear, block cap, background-work deferral, and resume.
• Session extras — project trust, plan mode, todos, checkpoints/rewind, AskUserQuestion, notifications, think keywords, session titles, /context, /init.

Slash commands: /init, /context, /goal, /memory, /todos, /rewind, /tasks, /agents, /plan, /plan-todos, /mcp, /hooks, and /output-style, alongside your own /dir:name commands, /skill:name skills, /plugin:name plugin commands, and each connected server's /mcp__server__prompt prompts. Where a name matches a Claude command, these differ: /init creates AGENTS.md where Claude creates CLAUDE.md; /memory edit opens a memory file in your editor (pi's externalEditor setting, VISUAL or EDITOR); /tasks lists background subagent runs only, not background shells. /todos and /plan-todos have no Claude counterpart.

pi has no general permission system, so most of what Claude routes through a permission prompt maps to hard behavior here: allowed-tools restricts the turn's tool set instead of pre-approving calls, and a hook that times out on PreToolUse or UserPromptSubmit fails closed. A hook's permissionDecision: "ask" is the exception: it shows a confirm dialog and lets the call through when you approve (a headless run has no dialog, so it blocks). An argument-scoped grant in an agent's tools: gives the agent the whole tool, as in Claude, where the scope does not limit the agent either. Claude still runs each call through its permission flow; pi has no such flow, so the call runs.

Trust is the other place pi-code is deliberately stricter. Claude states that "a claude -p run never shows the trust dialog" and loads the project's hooks, MCP servers, agents, commands, skills and rules anyway. pi-code refuses instead: with no stored decision and no UI to ask, a headless run in a project you have not already trusted loads none of them. A repository would otherwise get to run its own hooks and MCP servers in any CI job that checks it out, with nobody present to decline.

CLAUDE.md itself needs no extension: pi loads CLAUDE.md / AGENTS.md context files natively (global + walking cwd to root). context-imports.ts adds what pi's loader lacks, the @import resolution and Claude's other memory locations, without re-injecting the base (see docs/claude-md.md). Setting CLAUDE_CONFIG_DIR relocates the entire home config scope (settings, commands, agents, skills, plugins, output styles, memory, and the user CLAUDE.md); a project's own .claude/ is a separate scope and is unaffected.

extensions/internal/ holds the shared modules pi's loader must not treat as extensions (each file's header says what it owns); only internal/ keeps them out of pi's extension scan.

Vendored bases (question, notify, status-line) come from pi's MIT example extensions (see LICENSE).

Development

npm install
npm run check           # biome + strict tsc + knip + vitest with coverage floors, the whole gate
scripts/e2e-smoke.sh    # headless deterministic smoke: wire payload against a dead-port model, no real model (also runs in CI)
scripts/e2e.sh          # quick smoke of the real pi TUI via tmux (needs a working model)
scripts/e2e-full.sh     # every README feature end to end, model turns included (5-15 min)
scripts/record-demos.sh # re-records demos/*.tape with vhs at low thinking

Extensions live in extensions/, tests in tests/. Install a local checkout with pi install ./pi-code, then /reload after edits. Development needs Node >=22.19.

Contributing

Issues and pull requests are welcome. See CONTRIBUTING.md for setup, the test and review conventions, and the release process, and the Code of Conduct for community expectations. Report security issues privately as described in SECURITY.md. Release notes are on the Releases page.

License

MIT.