AGiXT — Extensible AI Agent Automation Framework

What it is

AGiXT is an open-source AI agent automation platform that combines 50+ LLM provider integrations, a plugin system for external tools, chain-of-thought workflows (called 'chains'), and persistent memory. It provides a web UI, REST API, and Python SDK for building and managing autonomous AI agents.

AGiXT is designed for teams that need customizable autonomous agents with tool integrations. It works with Claude, GPT, Gemini, Mistral, Ollama, and 45+ more providers, and can be set up in under 10 minutes with Docker.

How it saves time or tokens

AGiXT reduces development time by providing pre-built integrations with over 50 LLM providers and a plugin system for external tools. Instead of writing custom code for each provider or tool, you configure agents through its web UI or API. The chain-of-thought workflow system lets you compose multi-step agent behaviors without manually orchestrating API calls. With an estimated token budget of 4,000 tokens per workflow run, AGiXT keeps costs predictable through configurable limits and provider switching.

How to use

1. Clone the repository and start AGiXT with Docker:

git clone https://github.com/Josh-XT/AGiXT
cd AGiXT
docker compose up -d

1. Open http://localhost:8501 for the web UI and configure your first agent with an LLM provider and plugins.

1. Create chains (multi-step workflows) by linking agent actions — each chain step can call different providers, tools, or sub-agents.

Example

# Using AGiXT Python SDK
from agixtsdk import AGiXTSDK

agixt = AGiXTSDK(base_uri='http://localhost:7437')

# Create an agent
agixt.add_agent(
    agent_name='researcher',
    settings={
        'provider': 'anthropic',
        'model': 'claude-sonnet-4-20250514'
    }
)

# Run a prompt
response = agixt.prompt_agent(
    agent_name='researcher',
    prompt_name='Think About It',
    prompt_args={'user_input': 'Summarize the latest AI safety papers'}
)
print(response)

Related on TokRepo

• Multi-agent frameworks compared — Compare AGiXT with CrewAI, AutoGen, and other multi-agent orchestration tools.
• AI agent tools — Browse more agent-building frameworks and utilities curated on TokRepo.

Common pitfalls

• Starting with too many plugins enabled slows agent response time. Enable only the plugins your workflow actually needs.
• Chain-of-thought workflows can loop indefinitely if exit conditions are not set properly. Always define a maximum iteration count.
• Mixing providers within a single chain can cause format incompatibilities. Test each chain step individually before composing them.