Skip to main content

Agentix SDK

Pure-Python agent framework for any LLM. From personal projects to enterprise deployments.

Agentix is an LLM-neutral, pure-Python agent framework designed to scale from a solo developer's side project to a production enterprise deployment. It provides full native support for Anthropic Claude — including extended thinking budgets — and OpenAI models with reasoning effort control, alongside Google Gemini, DeepSeek, and any OpenAI-compatible endpoint. Built-in skills handle Word, Excel, PowerPoint, and PDF documents out of the box. A multi-channel messaging gateway connects agents directly to Slack, WhatsApp, Email, and HTTP webhooks, with no additional infrastructure required.

What is Agentix?​

Agentix is a production-ready agent harness SDK. It provides:

  • Single entry point — AgentixAgentOptions → AgentixClient
  • Multi-provider support — Anthropic, OpenAI, Gemini, DeepSeek, OpenAI-compatible
  • Built-in tools — Bash, Read, Write, Edit, Glob, Grep, WebSearch, WebFetch, and more
  • MCP integration — stdio, SSE, HTTP, JSON-RPC, and in-process SDK servers
  • Session management — persistent multi-turn conversations with automatic compaction
  • Hooks & permissions — full lifecycle control at every step
  • Plugin system — drop-in hook bundles with hash verification
  • Multi-agent orchestration — coordinator / sub-agent patterns with the Task tool
  • Structured output — JSON schema–constrained responses
  • Observability — OpenTelemetry tracing + structured JSON logging
  • Resilience — circuit breakers, configurable retries, backpressure
  • Gateway — expose agents on Slack, WhatsApp, Email, webhooks, and more
  • Effort / thinking budgets — effort="high" maps to extended thinking token budgets (Anthropic extended thinking)

Quick install​

Requires Python ≥ 3.11.

pip install agentix                  # all LLM backends included (Anthropic, OpenAI, Gemini, DeepSeek)
pip install agentix[redis]           # + Redis session storage & circuit breaker
pip install agentix[gateway]         # + Slack, WhatsApp, Email, and webhook adapters
pip install agentix[all]             # + Redis, all gateway adapters, and dev tools

30-second example​

import asyncio
from agentix import AgentixAgentOptions, query, ResultMessage

async def main():
    options = AgentixAgentOptions(
        name="my-agent",
        provider="anthropic",
        model="claude-sonnet-4-20250514",
        system_prompt="You are a helpful assistant.",
    )
    async for msg in query("What is 2 + 2?", options=options):
        if isinstance(msg, ResultMessage):
            print(msg.result)  # "4"

asyncio.run(main())

For multi-turn sessions, use AgentixClient directly:

import asyncio
from agentix import AgentixAgentOptions, AgentixClient, AssistantMessage, TextBlock

async def main():
    options = AgentixAgentOptions(system_prompt="You are a helpful assistant.")
    async with AgentixClient(options) as client:
        while True:
            prompt = input("You> ")
            async for msg in client.query(prompt):
                if isinstance(msg, AssistantMessage):
                    for block in msg.content:
                        if isinstance(block, TextBlock):
                            print(block.text)

asyncio.run(main())

Next steps​

  • Installation — detailed install instructions and optional dependencies
  • Quick Start — one-shot queries, multi-turn sessions, and custom tools
  • Tools & Permissions — built-in tools and permission callbacks
  • Hooks — lifecycle hooks for logging, modification, and control flow
  • MCP Servers — Model Context Protocol integrations
  • Multi-Agent — coordinator / sub-agent orchestration