Skip to content

SHARED CONTEXT. COLLECTIVE INTELLIGENCE.

Your AI agents.
One shared
direction.

Give every agent the same project context, memory and rules. Keep your whole team moving together.

Shared memory, rules and visibility for teams working with Claude Code, Codex and Cursor.

Early access
02 / INSTALL
Works with your team’s tools

Claude Code Codex Cursor

Knowledge stays.
Your team moves forward.

Your team. A shared direction.

01 / SHARED UNDERSTANDING

The same context.
A stronger team.

Your tools can work independently.
Your team’s knowledge shouldn’t.

Project memory, engineering conventions and session visibility your AI agents inherit on every run.

Illustrative example
Explore how CrewKit works
THE WORKFLOW01 — 04

One command.
A team that remembers.

From your first session to your next breakthrough. Everything you learn becomes part of what you build.

One CLI command connects Claude Code to your team's context.

Explore the documentation
CrewKit / checkoutIllustrative workflow
YOUR TERMINAL. CONNECTED.
~/acme/checkoutmain
crewkit code
brew install karibew/tap/crewkit
WORKSPACE RESOLVEDYou're right where you belong.acme / checkout / main
One command.Your whole team behind it.

One command. You're in.

Install CrewKit, then run crewkit code. Sign in inline — your organization and project resolve from the git remote.

03 / THE PLATFORMBUILT TO WORK TOGETHER

The work moves forward.
The knowledge stays.

Memory, collaboration, and standards that travel with your team. Evidence to make the next session better.

Governance for AI coding agents: shared agents, rules, playbooks, session analytics and experiments.

PROJECT MEMORY

Your next question.
Already in context.

PRDs, decisions, and past sessions become searchable project memory. Relevant context meets your agent at the start of every session.

  • Documents, transcripts, and prior sessions
  • Google Drive import and confirmed Slack capture
  • Privacy controls for sensitive and private work
Why did we choose optimistic locking?
FOUND IN YOUR PROJECT01
The decision was made.
The reasoning stayed.

Optimistic locking avoids contention on hot order rows — decided in the Q2 planning call.

Q2 planning transcriptProject artifact · acme / checkout
Illustrative project knowledge
COLLABORATION

A teammate.
With the backstory.

@mention a project-scoped agent in Slack. Get an answer in the thread, grounded in your team's context. Share sessions, annotate the work, and see who's doing what.

# eng-checkout PROJECT CHANNEL
M
Maya 10:02

@checkout-agent Why optimistic locking on order rows?

checkout-agent AGENT

From the Q2 planning call: it avoids contention on hot order rows.

Q2 planning transcript
FEBERVSeparate sessions.One shared view of the work.
SHARED CONVENTIONS

Your standards.
In every session.

Organization, project, and agent instructions work together. Sync playbooks, adapt behavior by role, and track convention adherence and overrides.

01
ORGANIZATIONEngineering playbook

Document decisions. Share the reasoning.

02
PROJECTCheckout conventions

Keep requests idempotent. Test retries.

03
AGENTReview with context

Apply the team's standards to this task.

Adapted to your role
CoachingCollaborativeAutonomous
rails-expert v2ILLUSTRATIVE DATA
Quality / session4.2/5+0.6 vs. v1
Average cost$0.029−14.7% vs. v1
Controlled experimentv1 → v2
Control 47 sessionsVariant 53 sessions

Compare quality, cost, and turns.
Check statistical significance before choosing.

Session → resource → exact version
OBSERVE & IMPROVE

Know what ran.
Learn what works.

Trace every session to its exact agent and skill versions. See costs, tokens, tool usage, and quality across your team — then test what should run next.

Visibility, down to the session.Team analytics, summaries, coaching insights, and resource attribution.
Experiments, backed by evidence.Controlled A/B assignments, quality comparisons, p-values, and version rollback.

One connected system.Built around your team's work.

Explore every feature
04 / THE MARKETPLACEBUILD YOUR CREWKIT

A head start.
Made your own.

Start with pre-built agents, skills, and playbooks. Shape them around your team, then sync them to everyone.

Curated agents, skills and playbooks for Claude Code, Codex and Cursor.

A SELECTION FROM THE LIBRARY

A specialist for the work ahead.

Domain expertise, ready to join your team.

01 / 03
  • AGENT01

    rails expert

    CrewKit
    rails-expert

    Rails conventions, migrations, testing

  • AGENT02

    frontend expert

    CrewKit
    frontend-expert

    React, TypeScript, accessibility

  • AGENT03

    API designer

    CrewKit
    api-designer

    REST/GraphQL API design patterns

  • AGENT04

    security reviewer

    CrewKit
    security-reviewer

    OWASP, auth, vulnerability scanning

Versioned Measured Ready to customize
YOUR TEAM'S CONTEXT

Start together.
Build your own way.

Build on shared resources with your organization's conventions and your project's context. Every resource change stays versioned, measured, and ready to roll back.

01PlatformShared foundation
02OrganizationYour team's standards
03ProjectYour working context

Find your starting point.Make it work like your team.

Explore the Marketplace
05 / YOUR NEXT CHAPTERBUILT TO MOVE TOGETHER
SHARED CONTEXTONE DIRECTION

Your next session.
A stronger starting point.

Bring your team's knowledge, standards, and experience into the work you're about to do.

Request Early Access

Early access is free. Request your invite by email.

Shared memory Team conventions Every version, traceable