API Reference / agent /

adapters/

The TuiML tools, for agent frameworks that do not speak MCP.

mcp covers clients that support the Model Context Protocol. These adapters cover the rest: if you are building an agent in Python rather than configuring one, they hand you the same tools in the shape your framework expects, with no MCP server involved.

Adapters

  • openai — OpenAI function-calling schemas.
  • anthropic — Anthropic tool definitions.
  • langchain — LangChain BaseTool objects.
  • crewai — CrewAI tools.
  • pydantic_ai — Pydantic-AI tools.
Each exposes get_tools(), returning every tool — the workflow tools plus tuiml_list and tuiml_describe, so an agent can discover what exists rather than being told in its prompt.

Installation

Each framework is an optional extra, imported lazily: 

pip install 'tuiml[openai]'        # or anthropic, langchain,
                                   # pydantic-ai, crewai
pip install 'tuiml[frameworks]'    # all five at once
python 
>>> from tuiml.agent.adapters.openai import get_tools   # doctest: +SKIP
>>> tools = get_tools()                                 # doctest: +SKIP

See Also

tools

The tool definitions all five adapters expose.
mcp

The MCP server, for clients that speak MCP.

Modules

anthropic

Anthropic adapter, TuiML tools in the Messages API tools shape....

0 classes · 3 functions
crewai

CrewAI adapter, TuiML tools as BaseTool subclasses....

0 classes · 2 functions
langchain

LangChain / LangGraph adapter, TuiML tools as StructuredTool objects....

0 classes · 2 functions
openai

OpenAI adapter, TuiML tools in OpenAI's function-calling shape....

0 classes · 3 functions
pydantic_ai

Pydantic-AI adapter, TuiML tools + a one-liner tuiml.agent.agent() agent....

0 classes · 3 functions