AI coding is powerful. Managing it shouldn't be guesswork.
crewkit is the institutional memory and governance layer for AI-assisted engineering. We give teams the memory, context, and governance to turn ad-hoc AI usage into a repeatable, optimizable practice.
The problem
Your team ships with AI agents every day — and the knowledge behind the code evaporates when each session ends. The productivity is real; the memory is not.
Engineering teams adopting these tools hit the same walls:
Context evaporates
Every decision behind the code disappears when the session ends — the meeting, the Slack thread, the reason you chose this pattern. The next session starts from zero.
Black-box sessions
No one knows what sessions cost, which agents succeed most often, or how junior developers are being guided. Data stays locked on individual machines.
Zero governance
Every developer has their own agent prompts, skills, and rules. There is no single source of truth, no role-based controls, and no compliance trail.
No way to improve
You can tweak a prompt, but you cannot measure whether the change helped. Without data, improving agent performance is guesswork — not engineering.
Why now
AI coding assistants went from experimental to essential in under 18 months. Engineering teams are spending real budgets, shipping real code, and scaling real workflows with AI — but the tooling to manage this shift does not exist yet.
Spend is scaling fast
Teams are going from a few hundred dollars to five and six figures in AI coding costs per month. Visibility is no longer optional.
Compliance demands governance
Regulated industries need audit trails, role-based controls, and enforceable coding standards — even when the code is AI-assisted.
Fragmentation is the default
Every developer configures agents differently. Without centralized management, consistency is impossible as teams grow.
Performance is unmeasured
No one can say which agent configuration produces the best results. The feedback loop that makes software engineering work is missing.
Our approach
crewkit is built around three pillars — remember, govern, collaborate — plus an improvement loop that turns ad-hoc AI usage into a managed, measurable practice.
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.
Improve
Every session records the exact version of every agent and skill that ran. Compare versions on cost, tokens, and quality scores; roll back what regressed.
What sets crewkit apart
Dashboards tell you what AI cost. crewkit changes what AI knows — and it lives where developers already work: the terminal.
Resource marketplace
A curated library of agents, skills, playbooks, and commands. Install proven configurations in seconds. Share what works across your organization.
3-tier inheritance
Platform, organization, and project-level configurations that compose and override cleanly. Change a skill once at the org level and it propagates everywhere.
Context-aware agents
Agents receive project context, team playbooks, and conventions automatically. Every session starts with the right knowledge.
Team memory
Institutional knowledge shouldn't live in one person's head. Docs, Drive imports, Slack threads, and past sessions become semantically searchable — and are injected at session start.
Data-driven optimization
Every resource version is tracked with performance metrics. A/B test configurations, measure impact, and deploy the version that works best.
Per-version attribution
“The agent got worse” stops being a feeling. Every session records the exact version of every agent and skill that ran; cost, tokens, and quality scores roll up per version.
How it works
Install in one command
brew, npm, choco, or curl. Run crewkit code: sign in inline, and your org and project resolve from the git remote.
Connect your context
Upload PRDs, transcripts, and contracts; import from Google Drive; capture Slack threads as project artifacts. Decisions stop evaporating.
Code with memory
crewkit code syncs your team's agents, skills, and playbooks, then injects the project context that matters. Mid-session, search everything the project knows — including prior sessions.
Improve on evidence
Every session records the exact version of every agent and skill that ran. Compare versions on cost, tokens, and quality scores; run A/B experiments with p-values; roll back what regressed.
Our vision
Every engineering team will use AI assistants. The question is whether they use them well. Today, most teams treat AI coding tools as individual developer utilities — no shared configurations, no measurement, no improvement loop.
crewkit exists to close that gap. We believe AI-assisted engineering should be a managed practice, the same way CI/CD transformed deployment and observability transformed operations. Teams that instrument their AI workflows will build faster, ship better code, and outpace those running blind.
We are building the platform that makes that possible — living project blueprints, searchable artifacts and past sessions, and agents that work as teammates, all inside the tools developers already use.
Built for the terminal
The crewkit CLI ships as a single fast native binary for macOS, Linux, and Windows — instant startup, no runtime dependencies.
Releases and the public issue tracker live on GitHub. The dashboard and API are hosted services designed for teams that need governance and analytics at scale.
Releases and issues on GitHub