A2Flow

A2Flow is a chat-driven workflow engine that rebuilds ITSM-style workflows around an AI agent.

A2Flow is a chat-driven workflow engine that rebuilds ITSM-style workflows around an AI agent. It plans work as a task graph, pauses for human approvals, and executes the rest. Procedures live as SKILL.md files in Git repositories, and the agent binds MCP tools scoped per task. It supports Gemini by default, with OpenAI and Anthropic models available via LiteLLM. The platform uses Docker Compose for self-hosting with PostgreSQL.

A2Flow at a glance

Best for: Software engineer, Product manager
Runs on: Web, Self-hosted
Pricing: Free
Released: November 2026
Category: AI agent tool
Website: kaitoy.github.io

Key features

Plans work as a task graph
Pauses for human approvals
Uses Agent Skills in Git
Integrates MCP tools scoped per task
Renders rich UI inside the chat
Supports Gemini, OpenAI, and Anthropic models

Use it for

Service requests
Change execution
Announcements

Alternatives to A2Flow

AI-Flow — Combine multiple AI models in one workflow and share them as simple apps.
OpenFlow — Build and run multi-agent AI pipelines on a visual canvas using real parallel agents.
A2A Protocol — Common language so agents can delegate work without sharing internals
Langflow — Builds and deploys AI agents visually with Python and major LLMs.
n8n — Self-hostable automation with code allowed in any step
Agenta — Build and automate AI agents that run on your own infrastructure or the cloud.
DeerFlow — Executes long tasks using sandboxes, memories, tools, and subagents.
Flowise — Modular building blocks and visual tools to build and scale agentic systems fast.
cascadeflow — Reduces agent inference costs by up to 90% with sub-5ms overhead.
Search2o — Builds model-neutral agents with strict spend limits and role-based access.
FastGPT — It uses hybrid retrieval and auto-cleansed data to prevent hallucinations.
Endo — Combine agents from different vendors into one thread and workflow.

All A2Flow alternatives