Cli
Checks your coding agent's edits against your team's decisions in tickets, chat, docs and git.
At a glance
Not tested yet- GitHub stars
- 24
- Weekly downloads
- 4,729
- Last commit
- 2026-11-19
- Licence
- MIT
- Transport
- stdio
- Auth
- none
npx -y @aligndottech/cli
What it does
Cli is the Align command line tool, packaged here as an MCP server. It collects your team's decisions from commits, tickets, chat and meetings into one graph, and re-reads those sources in the background. Your AI agent can query that graph over MCP to see the decision behind the code and where it came from. It is for teams whose agents know the code but not the company.

Key features
- Builds a decision graph from commit history, ADRs and RFCs, decision logs,
CLAUDE.md,AGENTS.md, tenets, architecture docs and recent merged pull requests. - Connects to GitHub, GitLab, Jira, Confluence, Linear, Notion, Slack, Teams and Zoom as read-only sources.
- An edit hook checks edits against the graph. It warns by default, and blocking is opt-in.
- On Claude Code the check runs before the edit is written. Cursor, Copilot CLI and Gemini CLI hand the result to the model after the edit lands.
- The agent can query the graph itself over MCP.
- In local mode the graph is a SQLite file at
~/.align/local.db. - Open source under MIT, with no account needed.
Quick start
Run it in a git repo. The first run builds a graph from your commit history and the repo's docs.
npm install -g @aligndottech/cli align
No Node? This installs a standalone binary instead.
curl -fsSL https://align.tech/install.sh | sh
To see it work with nothing installed and no account or AI key:
npx @aligndottech/cli demo
Add it to Claude Code from a terminal:
claude mcp add cli -- npx -y @aligndottech/cli
Or add this to .mcp.json in your project root (Claude Code) or .cursor/mcp.json (Cursor):
{
"mcpServers": {
"cli": {
"command": "npx",
"args": [
"-y",
"@aligndottech/cli"
]
}
}
}Environment variables
DO_NOT_TRACK: set it to turn off anonymous usage counts.
Example prompts
- Why does this error handler only keep status codes from 400 to 599?
- Before I change this file, check whether a past decision rules it out.
- Which decision introduced this line, and who made it?
What to watch out for
- Beta, pre-1.0.
- The npm install needs Node 22.19+.
- The
curlinstall line needs a POSIX shell on Windows (Git Bash or WSL). - In local-only mode the only thing sent to Align is anonymous usage counts, never code or decision text, unless you run
align share. - Tokens you paste for a connector go in your OS credential store, or in
config.jsonif there is none. - Telling replaced from conflicting decisions needs an AI key locally, or a team graph.
- Anthropic does not let other apps use a Claude subscription.
- The first demo run downloads a small on-device model (about 50 MB) once.
- Sources refresh only when you start
align, at most every 15 minutes per source. There is no timer. - Codex installs the plugin without trusting its hooks, so the check does nothing until you approve it in the hook review.
Not tested by tools4agents yet. Read the source repository before you install it.
Similar tools
Everything
The official test server. Exercises every part of the protocol, so it is the fastest way to check your setup works at all.
Filesystem
Gives an agent scoped read and write access to directories you name explicitly. The usual first MCP server people install.
Git
Read, search, and manipulate local Git repositories. Lets an agent inspect history and diffs without shelling out.
Sequential Thinking
Gives a model a scratchpad for multi step reasoning, letting it revise earlier steps instead of committing to a first answer.