Python Documentation

Quick Start Install Ax and run a typed LLM program.

How Ax Fits Together The Ax stack on one page: signatures, ai(), ax(), flow(), agent(), optimizers, and the AxIR compiler — and how they compose.

Advanced Start Follow the Ax story through selected runnable examples.

FAQ Plain-English answers to common Ax questions, plus a glossary: DSPy, GEPA, ACE, RLM, PEEK, AxIR.

Overview One agent harness at every size: typed signatures, tools, and a grounded runtime session.

Learn to build reliable AI workflows and agents A hands-on Python course for building dependable AI features, multi-step workflows, tool-using agents, and production automation.

Overview Provider-backed runnable examples generated from source headers.

DSPy The Ax take on DSPy-style typed LLM programs.

Signatures Typed input and output contracts for prompts, schemas, tools, and validation.

Tools Typed host functions, fn(), namespaces, function groups, and discovery.

MCP Connect MCP servers as Ax tools for generation, agents, prompts, resources, and runtime workflows.

MCP Subscriptions Discover an MCP endpoint catalog, select resources, maintain subscriptions, and wake or resume Ax programs safely.

Event Runtime Wake and resume Ax programs safely from subscriptions, webhooks, timers, queues, and application events.

Optimization GEPA, Pareto tradeoffs, demos, artifacts, and evaluation.

Playbook Evolving context playbooks that grow offline from examples and online from feedback.

LLMs Provider clients, model catalogs, routing, audio, embeddings, thinking, and context caching.

Telemetry Traces, metrics, usage, costs, optimizer metrics, and production debugging.

ai() LLM Models Create provider clients and select models.

ax() Generation Run structured generation with tools, streaming, validation, traces, and usage.

s() Signatures Parse or build signatures.

agent() Agents Build agents with tools, child agents, runtime profiles, and context controls.

flow() Workflows Compose typed programs into multi-step workflows; author or export them as mermaid flowcharts.

optimize() GEPA Tune programs with the top-level GEPA helper.

Overview Install language-specific Ax skills for coding agents.

ai() API Provider, model, routing, media, and embedding APIs.

ax() API Structured generation APIs.

s() API Signature and field APIs.

agent() API Agent, runtime, context, skill, memory, MCP, and RLM APIs.

flow() API Flow builder, workflow graph, and mermaid APIs.

optimize() API GEPA optimizer, Pareto, artifact, and evaluation APIs.

Python Full API Reference Generated Python API reference.

Agent Internals This page moved to the Agents section.

Agents This page moved to the Agents section.

Pages