The self-evolution layer for AI agents.
Agents that get better with every run — auditable evolution powered by the Genome Evolution Protocol (GEP): Genes, Capsules, and Events.
Website · Evolver · Developers
Most AI agents start every task from zero. EvoMap makes them learn: each run distills what worked into reusable Genes (strategy cards) and Capsules (case traces), recalls them on the next relevant task, and records the outcome as an Event — a transparent, auditable evolution loop instead of a black box.
| Concept | What it is |
|---|---|
| 🧬 Gene | A distilled, reusable strategy — selected by signal relevance |
| 💊 Capsule | A case trace of a successful run, recalled as cold memory |
| 📈 Event | An immutable record of what was tried and how it scored |
Engine & SDKs
- evolver — the GEP-powered self-evolving engine ⭐ 8.7k
- gep-sdk-js — JS SDK: signal extraction, gene selection, memory graph
- gep-mcp-server — MCP server exposing evolution tools to any MCP client
Editor & agent plugins
- Claude Code · Cursor · Codex · OpenCode
Convert anything into GEP assets
Curated knowledge
- awesome-agent-evolution ⭐ 145 · awesome-agent-swarm ⭐ 30
🌐 evomap.ai · 👥 Developer community