OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Evolve

Evolve is an open-source Python system for on-the-job learning in AI agents. It learns from agent task trajectories and provides memory for later iterations, with an MCP server for tool integration, vector storage, and LLM-based conflict resolution.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Evolve — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Consider Evolve when building an agent workflow where retaining experience across repeated tasks could help reduce repeated mistakes or make successful behavior more consistent.

Who it is for

  • Developers building AI agents that repeat tasks across sessions
  • Teams exploring memory and reusable guidance in MCP-based workflows
  • Builders evaluating agent learning alongside Claude Code or Codex

Documented uses

  • Retain lessons from agent task trajectories for later iterations
  • Retrieve task-relevant guidance through MCP tools
  • Choose between a simple filesystem memory backend and optional semantic vector-search backends
Before you choose

Requirements & limits

The project is open source under the Apache-2.0 license. Its README specifies Python 3.12 or higher and installation through uv or pip.

  • The default filesystem backend offers text matching; semantic similarity search calls for an optional pgvector or Milvus backend.
  • The README describes MCP integration and links Claude Code and Codex quick starts, but the high-level description does not provide integration steps or establish compatibility with a particular agent.
  • The project identifies itself as on-the-job learning from trajectories, distinguishing its purpose from a general-purpose agent runtime.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Evolve and start its MCP server

Install the Evolve MCP server and launch its documented Web UI and MCP service.

Before you start

  • Python 3.12 or higher
  • uv or pip
  • An OpenAI API key for direct OpenAI usage

Steps

  1. Install from PyPI with the command below, or follow the README's source-install steps using uv.

    Documented command
    pip install altk-evolve

  2. For direct OpenAI usage, set the OPENAI_API_KEY environment variable.

  3. Run the command below to start the Web UI and MCP server.

    Documented command
    uv run evolve-mcp

  4. Open http://127.0.0.1:8000/ui/ for the Web UI.

Expected outcome

The MCP server and local Web UI are started using the documented command.

Guide limitations

  • The default filesystem backend uses text matching; semantic vector search requires an optional backend such as pgvector or Milvus.
  • The README also documents a separate direct-uvicorn route for starting the Web UI without the MCP server's stdio process.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
AgentToolkit/altk-evolve · Official homepage title · retained in its source language
How we review sourcesSuggest a correction