crewkit

crewkit is the institutional memory and governance layer for AI-assisted engineering: agents work as teammates, prior work is one search away, and every session is attributed to the exact agent version that ran it.

Three Pillars

Remember — PRDs, transcripts, Slack threads, and past sessions become searchable project memory, injected automatically when a session starts.

Govern — Playbooks and conventions with an auditable challenge log, role-based autonomy, and privacy-first detectors: no code egress, no LLM in the loop.

Collaborate — Agents get a Slack handle, teammates see who's working on what live, and any session is one share link away.

How it works

1. Install the CLI and connect to your account
2. Connect your context — upload docs, import from Drive, link Slack, so decisions stop evaporating
3. Run crewkit code in any project — it authenticates, detects your project, syncs your team's agent configs, and launches Claude Code
4. Session data flows to your dashboard — tokens, costs, tool usage, conversation summaries
5. Iterate — tweak agent prompts, run A/B experiments, roll out what works

# Install
brew install karibew/tap/crewkit

# Start coding
cd my-project
crewkit code

The TUI wraps Claude Code with a real-time sidebar showing session metrics, git context, and AI-generated summaries.

Who is this for

Teams using Claude Code who want:

• Shared project memory — docs, transcripts, and past sessions searchable and injected at session start
• Consistent agent configs across the team (no "works on my machine" for AI prompts)
• Visibility into what AI is doing across all developers
• Guardrails — playbooks and conventions that enforce team standards
• Data-driven improvement — measure which agent versions perform best

Next steps

• Install crewkit
• Quickstart: zero to session
• Configure your agents