Open research lab for next-gen autonomous agents.
We explore the core capabilities that define the next generation of AI agents: self-organizing, self-verifying, and self-evolving — and build open-source reference implementations to prove them out.
| Direction | Question | Project |
|---|---|---|
| Self-Organizing | How do agent teams decompose work, coordinate, and resolve conflicts autonomously? | cc-manager |
| Self-Verifying | How can agent output prove its own correctness without human review? | behavior-driven-testing |
| Self-Evolving | How do agent systems improve from their own execution history? | cc-manager evolution pipeline |
- cc-manager — Research platform for autonomous agent orchestration. Parallel agents in git worktrees, self-evolution from execution data, proof-first merge pipeline.
- agent-ready — Codebase readiness scoring for autonomous agents. Beyond instructions — measurable operability standards.
- behavior-driven-testing — Verification discipline for agent-generated changes. Acceptance proof, regression safety, evidence trails.
- TeamClaw — Multi-agent team platform for cross-project coordination with shared memory and role-based access.
- LabClaw — AI-native scientific lab platform. First vertical application of Agent Next infrastructure, proving autonomous agent teams in experimental science.
Current-generation agents follow instructions. Next-generation agents learn, verify, and improve autonomously. The bottleneck is shifting from writing code to trusting code — and the infrastructure for that trust doesn't exist yet. We're building it.
Website · Roadmap · Contributing
Maintainer: @robotlearning123
