Evolver is a self-evolving engine for AI agents that uses Gene Expression Programming to enable autonomous capability development with full auditability.
The tool addresses the challenge of building AI agents that can improve their own behavior over time while maintaining transparency about how and why they evolved. It works by organizing agent capabilities into Genes, which are executable units that can be combined and modified, and Capsules, which are self-contained skill modules. The system tracks evolution through Events, creating an auditable record of every change an agent makes to itself. This approach allows agents to experiment with new skill combinations and prompt strategies without requiring manual retraining or redeployment.
Evolver suits teams building autonomous agents that need to demonstrate how their systems arrived at current behaviors, particularly in regulated environments or applications where explainability matters. It works well for projects where agents must adapt to new tasks by composing existing skills in novel ways, or where you want agents to optimize their own prompts and decision logic. The framework integrates with the Agent Protocol and supports memory systems and skill libraries, making it applicable to multi-agent scenarios and complex autonomous workflows.
The project shows active development with regular commits across core engine components, the CLI tooling, and example implementations. Work spans both foundational architecture updates and practical features like MCP integration and skill library expansion. The codebase receives consistent refinement to the Gene expression system and event tracking mechanisms that form the project's core. Documentation and example agents are maintained alongside the engine itself, indicating attention to developer onboarding.
