# Evo AI

> Open-source, self-hosted workbench for building AI agents: LLM, A2A, sequential, parallel, task and LangGraph workflow types with MCP tools and encrypted API keys.

- **Category:** AI Agents
- **Licence:** Apache-2.0
- **Latest version:** 0.1.0
- **Open source:** oss
- **Languages:** Typescript, Python
- **Platforms:** Docker, Docker Compose, Build from Source
- **Tags:** Mcp, Ai Agents, Observability, Agentic Ai, LLM, Next.js, AI, Python, Fastapi, Agent, Multi Agent, Langgraph, Crewai, Agent orchestration, Agentic workflow, A2a Protocol, Adk, Agent Workflows, Task Agents

## Description

Evo AI gives teams one web application for building, wiring up and running AI agents instead of scattering them across scripts. An agent can be a single language-model agent or a composite one: sub-agents run in sequence, in parallel, in a bounded loop, as a one-off task, or inside a custom graph built with LangGraph, and agents can also speak Google's A2A protocol to talk to other agents. Alongside agents you manage clients, MCP server connections, custom tools and folder groupings, while model credentials are kept encrypted; optional Langfuse tracing streams agent runs over OpenTelemetry. Human access is email-and-password with JWT tokens, verified addresses, a recovery flow and lockout after repeated failures. Self-hosting means a FastAPI backend plus a Next.js frontend: the compose stack pulls the API image with PostgreSQL and Redis, a second compose file starts the web container on port 3000, and the README also covers a make-based install with Alembic migrations and seed data. Release 0.1.0 arrived in May 2025 and the repository shows no pushes after June 2025; the CrewAI engine alternative to Google ADK is still in development, and the vendor sells hosted monthly plans through its site.

## Links

- [Homepage](https://github.com/evolution-foundation/evo-ai)
- [Website](https://evo-ai.co)
- [Docker](evoapicloud/evo-ai)

---
Source: https://selfhost.directory/project/evo-ai