fastllm-claude-code

Use an authenticated Claude Code CLI as a FastLLM provider. The adapter runs models through fastclaude and supports streaming, non-streaming calls, and client-owned tool loops.

Your application executes tool requests and supplies the results in the next call. FastLLM replays canonical history after each tool round. fastclaude continues that history in a fresh process.

Usage

Installation

Install from PyPI:

pip install fastllm_claude_code

Or install from conda:

conda install -c AnswerDotAI fastllm_claude_code

For the latest source version:

pip install git+https://github.com/AnswerDotAI/fastllm-claude-code.git

Documentation

See the API documentation and source repository. Package listings are on PyPI and conda.

How to use

Installing the package registers the claude_code transport through FastLLM’s provider entry point. The Claude CLI must already be installed and authenticated for the current user.

Use the provider prefix in a FastLLM call:

from fastllm.acomplete import acomplete
from fastllm.types import sonn

answer = await acomplete('Answer briefly: what is 2+2?', model=f'claude_code/{sonn}')

For a client-owned tool loop, pass standard Responses API or Chat Completions function schemas. A response containing tool calls ends the turn. Execute the requests in your application, extend the history with their results, and call again.

Calls do not return a response id. Supply complete history each time and reuse prompt_cache_key to retain Claude’s account-context reminders:

first = await acomplete(messages, model=f'claude_code/{sonn}', tools=tools, prompt_cache_key='my-dialog')
assert first.tool_calls and first.response_id is None

final = await acomplete(messages_with_results, model=f'claude_code/{sonn}', tools=tools, prompt_cache_key='my-dialog')

Set stream=True for FastLLM’s normalized async stream. Non-streaming calls collect the same stream into one Completion. The adapter never executes tool requests.