THE IDEA BEHIND CREWKIT
Good work
should leave
a memory.
The session ends. The knowledge shouldn't. We're building the shared memory and governance layer for teams that engineer with AI.
Why we're hereGreat teams don't
start from zero.
Neither should their AI.
Your team ships with AI agents every day. But the reasoning behind the code can disappear when a session ends: the decision in a meeting, the context in a Slack thread, the lesson from a previous attempt.
CrewKit exists to close that gap. We believe AI-assisted engineering should be a shared, measurable practice, with living project blueprints, searchable knowledge, and agents that work as teammates.
02 / HOW WE THINK
A few things
we build on.
More context. Clearer expectations. A crew that gets better together.
See it in the product01Shared memoryRemember the why.
The code tells part of the story. Project documents, transcripts, Slack threads, and past sessions carry the decisions behind it. CrewKit makes that context searchable, so the next session can build on the last.
Living project blueprints keep the plan alongside the work. Relevant project knowledge joins the session where it is needed.
02Thoughtful governanceMake good judgment repeatable.
A team's conventions should be shared knowledge. Playbooks, role-based autonomy, and an auditable challenge log give people and agents a common way to work.
Platform, organization, and project configurations compose through three levels of inheritance. The shared foundation can evolve without losing the context of an individual project.
03Real collaborationWork as a crew.
AI-assisted engineering is a team practice. Agents participate in the workflow, teammates see active work, and sessions can be shared instead of staying locked on individual machines.
Connect the people, agents, and resources around a project. Shared context helps everyone move in the same direction.
04A feedback loopImprove on evidence.
Every session records the versions of the agents and skills that ran. Cost, tokens, and quality signals make changes something you can evaluate, instead of something you can only feel.
Compare resource versions, run experiments, and roll back a regression. The learning should be as continuous as the work.
03 / CLOSE TO THE WORK
At home in
your terminal.
The CrewKit CLI is a fast native binary for macOS (Apple Silicon) and Linux. On Windows, run the Linux build under WSL. Your team's agents, skills, and playbooks meet you where you already work.
The dashboard and API bring shared context, governance, and analytics to the whole team.
Releases & issuescrewkit code04 / LET'S TALK
Good things start
with a conversation.
We'd love to hear what you're building,
what's working, and what could be better.