April 5, 2026·3 min·Nima Nejat
axint is on PyPI.
The Axint Python SDK is live. define_intent() in Python compiles to the same Swift output as TypeScript. pip install axint.
Short one. We shipped the Python SDK today.
bash pip install axint
```python from axint import define_intent, param
send_message = define_intent( name="SendMessage", title="Send Message", description="Sends a message to a contact", params={ "recipient": param.string("Who to message"), "body": param.string("Message content"), }, ) ```
bash axint compile intent.py --out Generated/
Same Swift output as TypeScript. Same validator. Same generator. The Python SDK produces the same [language-agnostic IR](/blog/intermediate-representation-design) as the TypeScript compiler — two front-ends, one backend.
The motivation was straightforward: the AI agent orchestration ecosystem lives in Python. LangChain, CrewAI, AutoGen — all Python. With axint, a Python agent can programmatically generate and compile App Intents without ever touching Node.
v0.1.0 covers define_intent() and the full param.* API. Entity support and the MCP bridge land in v0.2.0. We [parse the AST statically](/blog/ast-parsing-vs-execution) instead of executing user code, which turned out to be a surprisingly interesting design decision.
Try it. pip install axint.