Your Autonomous AI Engineering Team.

Stop wrestling with generic chat assistants. aiagentflow is a local-first, multi-agent framework built specifically for software engineering. Architect, code, review, and test—all from your terminal.

bash
How it works

A Deterministic Multi-Agent Pipeline

We replaced chaotic LLM loops with a strict Directed Acyclic Graph (DAG) of specialized engineering personas. Predictable inputs, locally verified outputs.

1

Analyze & Plan

The Architect Agent analyzes your codebase and drafts a solid implementation plan before a single line of code is written.

2

Smart Coding

The Coder Agent executes the plan, writing clean, maintainable code across your entire stack.

3

Automated Review

The Reviewer Agent audits the code for bugs, security vulnerabilities, and style guide adherence.

4

Test & Validate

The Tester Agent writes unit tests while the Fixer Agent automatically resolves any review comments or failing tests. Finally, the Judge ensures all requirements are met.

Why aiagentflow?

Built for Autonomous Engineering

Everything you need to ship software faster, directly from your terminal. No cloud lock-in, no data leaks, no context exhaustion.

Local-First

Your code stays on your machine. No sending proprietary architecture to third-party servers.

Framework Agnostic

Works with React, Vue, Node, Python, and more. If you can build it, our agents can orchestrate it.

Reliable Outputs

By injecting your existing architecture docs, API specs, and PRDs as direct context, we eliminate hallucinations and align the output to your specific coding standards.

Self-Healing

Built-in Fixer, Judge, and Tester agents automatically detect, retry, and resolve errors before you even see them.

Extensible

Write custom agents or tweak prompts to match your team’s specific coding standards and workflow.

Streaming & Auto-Detection

Watch agents code in real-time with --stream. Automatically detects available models from Anthropic, OpenAI, Gemini, and Ollama.

Use your favorite AI

Compatible with the best models

A

Anthropic (Claude)

Ideal for complex architectures. Optimized for claude-sonnet-3.5 and above.

O

Ollama (Local Models)

Total privacy and zero cost. Run Llama 3 or DeepSeek directly on your machine.

O

OpenAI (GPT-4o)

Pure speed for task execution and smart agent routing.

G

Google Gemini

Leverage its massive context window to analyze entire repositories in a second.

+ More providers like Mistral coming soon

Quick Demo

One tool,infinite possibilities

Whether integrating into your workflow or via CLI, aiagentflow adapts perfectly to your way of working.

Quick Demo

User Intent

"Build a responsive Next.js header"

$aiagentflow run "Build a responsive Next.js header" --context docs/design-system.md
ReadyUTF-8
Line 1, Col 1Zsh
Get Started

Up and running
in seconds

Add @aiagentflow/cli to any Node.js project. Our zero-dependency engine ensures your environment stays clean.

$ npm install @aiagentflow/cli

OpenAI

GPT-4o, GPT-4 Turbo

Anthropic

Claude 3.5 Sonnet, Opus

Google

Gemini 1.5 Pro, Flash

Ollama (Local)

Llama 3, Mistral

Open Source

From the community, for the community

aiagentflow is an MIT-licensed project. Join hundreds of devs shaping the future of autonomous engineering.

1.2k+
Stars on GitHub
45+
Contributors
Discord
Community Hub
Make your first contribution
Documentation | Bug Fixes | Feature Requests