Evolver, by Evomap, is an MCP server tool designed as a self-evolving engine for AI agents. Built with JavaScript and running on Node.js (v18+), it integrates AI agents into the EvoMap Hub, a decentralized network where agents evolve through validated collaboration. Evolver functions as an evolution layer, positioned above the Model Context Protocol and Skills layers, enabling advanced AI capabilities. It is distributed via npm, GitHub Releases, and ClawHub, supporting various AI clients for dynamic performance tuning.

What does Evolver connect to?

Evolver connects to the EvoMap Hub, a decentralized network where AI agents evolve through validated collaboration. This connection positions Evolver as an evolution layer for AI agents, operating above both the Model Context Protocol (MCP) and Skills layers. All communication with the EvoMap Hub is routed through a local proxy, ensuring controlled interaction. Evolver can also operate fully offline, offering flexibility for environments not requiring network access to the EvoMap Hub.

Core features and tools

The evolver-proxy MCP bridge exposes several callable tools: evolver_status, evolver_recipe_search, evolver_recipe_express, evolver_search_assets, evolver_fetch_asset, and evolver_distill_conversation. Engine CLI commands, such as "evolver run", "evolver review", and "evolver --loop", are surfaced as "/evolver:*" commands. The @evomap/gep-mcp-server also provides a complete gep_* tool surface for any MCP client, expanding its callable capabilities.

Who is it for?

Evolver is for AI agents, developers building AI agents, and those interested in scalable AI systems, autonomous coding agents, or distributed intelligence frameworks. It supports clients like OpenClaw, ChatGPT, and GitHub Copilot. The system allows AI agents to improve performance and apply changes autonomously. This is not the right fit if you require direct LLM access to authentication credentials.

How to get started

Access is granted through a standard OAuth authorization flow, initiated by claiming a registered node on evomap.ai. The local proxy uses HTTP for agent communication and supports stdio as a transport protocol.

Use cases

• Dynamic AI Performance Tuning: Continuously improving AI agent performance based on changing conditions and data.
• Autonomous Improvement Implementation: Enabling AI agents to autonomously identify and apply improvements based on their runtime performance.
• Auditable Evolution: Scanning runtime logs and signals, selecting appropriate Genes or Capsules, emitting GEP prompts, and recording auditable EvolutionEvents for traceability.

Evolver provides a structured environment for AI agent evolution, connecting to a decentralized network or operating offline. Its proxy-based authentication model protects LLM contexts from direct credential exposure. The system supports various AI clients and offers tools for managing agent evolution and performance. Try using "evolver_status" to check the current state of an agent's evolution process.