MCP.so
ログイン

Align CLI MCP Server認証済み注目

@aligndottech

基本情報

Align CLI MCP Server について

Your AI agents know the code. They don't know the company. A local decision graph from your git history, checked by your coding agent before it edits.

Align CLI MCP Server の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "align": {
      "command": "npx",
      "args": [
        "-y",
        "@aligndottech/cli",
        "mcp"
      ]
    }
  }
}

Align CLI MCP Server のツール

ツール一覧はまだありません

この掲載には表示できるツール一覧がまだありません。サーバーにツールがないという意味ではありません。対応機能はプロジェクトのドキュメントをご確認ください。

上のボタンで一時的なサンドボックスでこのサーバーを実行し、ツール一覧を取得できます。

概要

What is Align CLI?

Your AI agents know the code. They don't know the company. Align builds a local decision graph from your repo's git history, merged PRs, ADRs and CLAUDE.md / AGENTS.md, and from Jira, Linear, Notion, Confluence, Slack, Teams and others you connect, all read-only. Your coding agent asks it what was decided and why, and every edit is checked against it before it is written. If an edit rewrites lines an earlier decision wrote, it stops and asks you first.

It runs on your machine. No account, MIT licensed.

How to use Align CLI?

  1. Install: npm i -g @aligndottech/cli (Node 22.19+), or curl -fsSL https://align.tech/install.sh | sh for a standalone binary.
  2. In a git repo, run align. The first run builds the graph from your history, then lists the files it will write (MCP config, hooks, a block in CLAUDE.md / AGENTS.md) and asks before writing them.
  3. Ask your agent something like "why did we switch to postgres?", or ask it to change code an old decision covers and watch it stop and cite the decision.

To wire an agent by hand, use the config on this page, or run align mcp --setup.

Key features of Align CLI

  • Decisions from where they were made: commits that state a reason, PR descriptions, ADRs and agent rule files out of the box, and team tools when you connect them.
  • Checked before the edit lands: edit hooks for Claude Code, Codex, Cursor, Copilot CLI, Gemini CLI, pi and OpenCode, plus MCP tools for any client.
  • Scoped to the repo you're in: pick which Jira project, Linear projects, Notion pages or Confluence page trees belong to each repo.
  • Your model, or none: inside your agent no key is needed. Align's own agent runs on your API key, a local Ollama model, or your installed Claude Code.
  • Local by default: the graph is a SQLite file on your machine and embeddings run on-device.

Use cases of Align CLI

  • Stop an agent from undoing a decision it never saw, such as a status-code rule set in an old PR.
  • Give an agent the reason behind a file before it edits it.
  • Keep one decision record across git, tickets, docs and chat instead of a CLAUDE.md that goes stale.

よくある質問

1

Does it send my code to Align?

No. In local mode your decisions and code stay on your machine. If you set an AI key, questions and proposed changes go to your own provider, not to Align. Full list: https://github.com/aligndottech/align-cli/blob/main/docs/local-mode.md

2

Which clients does it work with?

Claude Code, Cursor, Codex, Copilot CLI, Gemini CLI, VS Code, Windsurf, Zed, Claude Desktop, pi and OpenCode.

3

Try it on a repo you didn't write

Clone fastify and ask your agent to make 1xx-3xx errors keep their status code in lib/error-status.js. It stops and cites fastify #7089, the PR behind that rule.

コメント

「AI とエージェント」の他のコンテンツ