MCP Servers · EvoMap

evolver

Evolver is a Node.js-based agent evolution engine that uses Gene Expression Programming (GEP) to automatically improve AI agent behavior. It encodes agent experience as reusable 'Genes' and 'Capsules' instead of ad hoc prompt tweaks, with full audit trails via Git integration.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/EvoMap/evolver

8.9k

GitHub stars

819

Forks

JavaScript

Primary language

GPL-3.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	EvoMap/evolver
Owner	EvoMap
Primary language	JavaScript
License	GPL-3.0 — OSI-approved
Stars	8.9k
Forks	819
Open issues	4
Latest release	v1.91.0 (2026-07-07)
Last updated	2026-07-07
Source	https://github.com/EvoMap/evolver

What evolver is

A GEP-driven self-evolution framework for autonomous agents that operates via CLI, generates evolution prompts from memory signals, integrates with major IDEs (Cursor, Claude Code, Codex, OpenClaw), and requires Node.js ≥18 and Git. Requires offline operation but offers optional EvoMap network hub integration.

Quickstart

Get the evolver source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/EvoMap/evolver.git2cd evolver3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Iterative AI agent improvement in Git-tracked codebases

Evolver shines when you need systematic, auditable evolution of agent prompts and behavior without manual trial-and-error. Works seamlessly in existing Git workflows.

Research into test-time adaptation and gene-based learning

Backed by peer-reviewed arXiv research (2604.15097) showing gene-based evolution outperforms skill-documentation approaches across 4,590 trials. Suitable for teams exploring procedural-to-strategy transitions in AI.

IDE-integrated agent development (Cursor, Claude Code, OpenClaw)

Tight integration via hooks enables real-time evolution prompts during active coding sessions. Particularly valuable in Cursor and OpenClaw environments where session context is rich.

Implementation considerations

Must run inside a Git-initialized repository; non-Git projects fail immediately with clear error.
Requires Node.js ≥18; no support for older LTS versions.
Memory folder structure (`./memory/`) is auto-created but location is relative to working directory; ensure consistent CWD across runs.
Gene/Capsule selection is built-in and offline; no external ML service dependency.
Review mode (`--review`) enables human-in-the-loop; `--loop` runs as daemon (background integration).

When to avoid it — and what to weigh

Non-Git or version-control-free projects — Evolver explicitly requires Git initialization and uses Git for rollback and blast-radius calculation. Will fail immediately in non-Git directories.
Projects requiring permissive open-source guarantees — License is GPL-3.0 (copyleft). Future versions announced as 'source-available' (not open source). Commercial derivative work requires GPL compliance or future commercial licensing.
Closed-source or proprietary agent systems — GPL-3.0 mandates source disclosure if distributed. Complex licensing implications if you cannot open-source downstream code. Requires legal review.
Offline-only, air-gapped environments without optional hub setup — While core engine runs offline, integration with EvoMap hub requires network access for skill sharing and leaderboards. Pure offline use is possible but feature-limited.

License & commercial use

Currently GPL-3.0-or-later (since 2026-04-09; initially MIT on 2026-02-01). All past and current versions remain fully open source and freely usable under their respective terms. Future releases announced as 'source-available' (not explicitly defined in data; specific license TBD).

GPL-3.0 requires that any distributed software incorporating Evolver must either open-source its code or obtain explicit exemption. Requires legal review before embedding in closed-source commercial products. The project's stated move toward 'source-available' licensing suggests future versions may introduce commercial license terms or restrictions. Internal/non-distributed use is GPL-compliant.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Needs review
Deployment complexity	Low
DEV.co fit	Strong
Assessment confidence	High

Security considerations

No explicit security audit or threat model stated in data. Project handles agent prompts, memory logs, and Git operations; ensure `.env` credentials (A2A_NODE_ID) are not committed to version control. GPL-3.0 code is publicly auditable. Offline-first operation limits external attack surface. Hub integration (optional) introduces network dependencies; vet evomap.ai security posture independently. No claim of security certification or penetration testing in provided data.

Alternatives to consider

LangChain / LangSmith (prompt versioning & observability)

Offers prompt versioning and runtime observability but is less focused on autonomous agent self-evolution and gene-based optimization. More suitable for general LLM application development than GEP-driven agent iteration.

OpenAI Evals / Anthropic Evaluations

Provide evaluation and benchmarking frameworks for LLM behavior but do not automate evolution loop or gene encoding. Require manual prompt tuning; suitable for validation rather than autonomous improvement.

AutoGPT / AgentGPT frameworks

Broader agent scaffolding; do not focus on systematic evolution via GEP or audit-trail-based self-improvement. More general-purpose; less specialized for iterative gene-based optimization.

Software development agency

Build on evolver with DEV.co software developers

Install Evolver via npm and run your first evolution in 30 seconds. Fully offline. No credit card required. Requires Node.js ≥18 and Git.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Evolve super-agent-party godot-mcp

Related on DEV.co

Explore the category and the services that help you build with it.

MCP Servers AI development and agent frameworks Custom software and prompt governance Talk to DEV.co

evolver FAQ

Can I use Evolver in a closed-source commercial product?

Not without legal review. GPL-3.0 requires source disclosure or commercial exemption. Future 'source-available' versions may introduce paid licensing. Consult your legal team before production use.

Does Evolver require internet/hub connectivity?

No. Core engine runs fully offline. Hub connection (evomap.ai) is optional and only unlocks network features (skill sharing, leaderboards, worker pool). Standalone use is fully supported.

What happens if I run Evolver outside a Git repository?

It will fail immediately with a clear error message. Git is a hard requirement for rollback and blast-radius calculation. Initialize a Git repo in your project directory first.

How does Evolver differ from traditional prompt engineering?

Evolver encodes agent experience as compact Genes (via GEP) rather than longer skill documentation. Peer-reviewed research (arXiv 2604.15097) shows genes are more robust and yield better performance gains; Evolver automates this evolution loop with audit trails.

Software developers & web developers for hire

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If evolver is part of your mcp servers roadmap, our team can implement, customize, migrate, and maintain it.

Ready to evolve your AI agent automatically?

Install Evolver via npm and run your first evolution in 30 seconds. Fully offline. No credit card required. Requires Node.js ≥18 and Git.

Talk to DEV.co Schedule a Consultation