rei

Headless CLI for Reikon — local-first git repository analysis and the terminal review loop.

It runs the same analyzers as the desktop app — health score, complexity, duplicate code, dead code, dependency audit, bus factor, commit quality — and brings the human↔agent loop to your shell: rei review and rei status work the flags and dispatch ledger, and rei mcp hosts Reikon's tools for a coding agent with no desktop app installed. No account, no upload — everything runs on your machine.

Install

npm install -g @reikondev/cli

Usage

rei analyze --path ./your-repo

Usage: rei <command> [options]

Commands:
  analyze           Analyze a git repository
  review            Work the review loop (list / reply / resolve / request-fix / send-back)
  status            Dispatch ledger — what's in flight across your repos
  mcp               Run the MCP server over stdio (for coding agents)

Options:
  --path, -p <dir>      Path to git repository (default: .)
  --format, -f <fmt>    Output format: json | html | summary (default: summary)
  --output, -o <file>   Output file (default: stdout for json/summary, reikon-report.html for html)
  --quiet, -q           Suppress progress output to stderr
  --no-cache            Skip cache, force fresh analysis
  --watch, -w           Re-run on every new commit
  --version, -v         Show version
  --help, -h            Show this help

Review loop

rei review and rei status bring Reikon's human↔agent review loop to the terminal — reading and mutating the same state as the desktop app and the MCP server, so a reply you make in the shell shows up in the desktop Review tab, and vice versa.

rei review                                # list in-flight flags + threads (--format json for agents)
rei review reply <id> "looks good"
rei review resolve <id>
rei review request-fix <id>               # the connected agent picks it up over MCP, fixes, returns
rei review send-back <id> "still leaks on the error path"
rei review --watch                        # live stream of flag/return changes

rei status                                # the dispatch ledger — returned / out / goals across repos
rei status --watch

Authoring goals and the visual substrate (ownership maps, hotspot charts) stay in the desktop app; the CLI is the loop's read + act surface.

Exit codes

rei analyze returns a real CI gate, in every output format (summary, json, html):

• 0 — health at or above the warn threshold
• 1 — below warn (warnings)
• 2 — below the fail threshold, or an analysis error

Set the thresholds in .reikon/config.json (see Config); they default to warn 70, fail 50. Wire it straight into a CI step — a non-zero exit fails the job:

rei analyze --quiet

Config

Reads .reikon/config.json from the target repo, exactly like the desktop app — teams, aliases, ignore paths, health weights, all apply the same way. One CLI-only addition:

{ "cli": { "warnBelow": 70, "failBelow": 50 } }

MCP server (for coding agents)

The CLI ships Reikon's MCP server, so a coding agent (Claude Code, Cursor, …) can pull project context — health, complexity hotspots, diff review, ownership — with no desktop app installed:

cd your-project
rei analyze                       # warm the context (optional — see below)
claude mcp add reikon -- rei mcp

rei mcp serves over stdio, scoped to the git repo it's launched in. rei analyze writes a cached context the server reads, so health/complexity tools answer instantly. Skip it and the first such call returns a "run rei analyze" hint and warms the context in the background (a retry shortly after succeeds); the diff and annotation tools work immediately either way, since they read live git. The CLI and desktop app share on-disk state, so annotations stay in sync between them.

Full docs

Flags, output formats, and CI integration examples: reikon.dev/docs.

• Install guide
• CI integration
• Full flag reference
• npm package

License

MIT