Vibe Coding DiscoverAWorld
View on GitHubSearch, understand, reproduce, and improve an idea with ease
Python agent harness that runs a multi-agent CLI (orchestrator, developer, evaluator) to build, evaluate and self-improve agents and apps. Includes skills, MCP support, deep-search browsing, RL training and benchmark harnesses like GAIA and BrowseComp.
Use Cases
Build and orchestrate multi-agent applicationsAutonomous app scaffolding with build-evaluate-evolve loopsDeep search and web research agentsOne-prompt video/image/audio generation workflowsAgent RL training and learning experimentsRunning benchmarks like GAIA, BrowseComp, xBench, BFCLv2Structuring large codebases with CAST for agent editingAuthoring and sharing reusable agent Skills
Built With
- Language
- Python
- Frameworks
- MCP · Docker · pytest
Tags
multi-agent · agent-harness · agent-framework · agent-runtime · cli · skills · self-evolution · evaluation-loop · context-management · deep-search · rl-training · benchmarks · code-generation · world-model · python