crewkit
FeaturesGitHub
Sign InGet Started

AI Agent Management Built for Developers

CLI-first platform for managing Claude Code agents across your organization. Role-based configs, A/B testing, and accuracy monitoring in one powerful toolkit.

curl -fsSL https://crewkit.io/install.sh | sh
Get Started Free
CLI-First
Built for the terminal
Team-Scale
Consistent configs
Open Source
MIT licensed CLI
terminal
$ crewkit code
🔍 Detecting project from git remote...
✓ Found organization: acme-corp
✓ Found project: customer-portal
📥 Fetching effective configs for senior role...
✓ Synced rails-expert.md
✓ Synced frontend-expert.md
✓ Synced api-designer.md
🚀 Launching Claude Code...

Everything you need to manage AI agents

Built for development teams who need control, visibility, and confidence.

Role-Based Agent Configuration

Automatically customize agent behavior based on developer experience level. Junior devs get coaching mode, seniors get full autonomy.

  • 3-tier config inheritance - Base agents, org overrides, project-specific context
  • Coaching mode for juniors - Agents guide rather than code, helping developers learn
  • Safe overwrites with backups - Never lose local changes, automatic checksums
.claude/agents/rails-expert.md
# Rails Expert
**Role detected: Junior Developer**
## Coaching Mode Active
CRITICAL: You are in COACHING MODE.
- DO NOT write code for them
- GUIDE them step-by-step
- ASK clarifying questions
- EXPLAIN concepts
## Organization Standards
- Use RSpec (not MiniTest)
- Follow Rubocop Shopify guide

A/B Test Your Agent Configurations

Test configuration changes with controlled experiments. Statistical analysis helps you deploy the best-performing agents.

  • Auto-generated experiment names - swift-amber-falcon, clever-jade-phoenix
  • Real-time metrics - Track accuracy, cost, and performance live
  • One-click deployment - Ship winning variants when statistically significant
terminal
$ crewkit experiments create rails-expert
✓ Created experiment: swift-amber-falcon
Control: Current config
Variant: Your modified config
$ crewkit experiments metrics swift-amber-falcon
Control Group:
Sessions: 47 | Accuracy: 87.2%
Avg Cost: $0.0342
Variant Group:
Sessions: 53 | Accuracy: 92.1%
Avg Cost: $0.0289
✓ Statistically significant (p=0.023)

Track Agent Performance & Costs

Track session history, token usage, and costs across your organization. See which agents perform best.

  • Session history - Track every coding session with full context
  • Cost tracking - Monitor token usage and API spend per project
  • Success rate tracking - See which agents and projects perform best
dashboard - session analytics
Organization: acme-corp
Period: Last 7 days
Sessions Overview:
Total Sessions: 147
Success Rate: 89.1%
Avg Duration: 12m 34s
Cost Breakdown:
Total Spend: $18.42
Avg per Session: $0.125
Top Agents:
rails-expert: 52 sessions (91% success)
frontend-expert: 38 sessions (87% success)
crewkit

AI agent management built for developers.

Product

  • Features
  • Documentation
  • Install CLI

Resources

  • GitHub
  • Releases
  • Report Issue
  • System Status

Legal

  • Privacy
  • Terms

© 2026 crewkit. All rights reserved.

Command Palette

Search conversations, projects, playbooks, and more