Tailrace - Agent data governance for TypeScript. Docs · All packages · @tailrace/core

@tailrace/cli

The tailrace binary: scaffold governed agent apps, add config to existing projects, scan for secrets, and enforce policy in Claude Code hooks.

# New project (bundled templates; no monorepo clone)
npx @tailrace/cli create next my-agent
npx @tailrace/cli create cloudflare my-agent
npx @tailrace/cli create openai my-agent

# Existing project
pnpm add -D @tailrace/cli
pnpm exec tailrace init
pnpm exec tailrace install-hooks   # merges PreToolUse/PostToolUse into .claude/settings.json

create is distinct from init: create writes a full starter app; init only adds tailrace.config.ts + .tailrace/config.json to the current tree.

Hook hot path loads .tailrace/config.json only (no TypeScript transpile). Spawn-to-exit p50 is gated under 150ms in CI.

Custom patterns (v2 config): add a recognizers array to .tailrace/config.json with validated regex sources. tailrace scan and tailrace hook load them via definePatternRecognizer. See Write custom recognizers.

| Command | Purpose | | -------------------------------- | -------------------------------------------------------------------- | | tailrace create <target> [dir] | Scaffold next / cloudflare / openai agent (--force, --install) | | tailrace init | Write tailrace.config.ts + .tailrace/config.json | | tailrace scan <path\|-> | Tier 0 scan; exit 1 on block-class entities (--json) | | tailrace install-hooks | Non-destructive Claude Code settings merge (--scope project\|user) | | tailrace hook | Claude Code hook handler (stdin JSON → stdout JSON) |

Docs: CLI reference