Skip to main content

hf-inference-acp

Hugging Face inference agent with ACP (Agent Client Protocol) support, powered by fast-agent-mcp.

Installation

uvx hf-inference-acp

What is this?

This package provides an ACP-compatible agent for Hugging Face Inference API. It allows you to use Hugging Face's Inference Providers through any ACP-compatible client (like Toad).

ACP sessions are always connection-scoped.

Features

  • Setup Mode: Configure Hugging Face credentials and model settings
  • Hugging Face Mode: AI assistant powered by Hugging Face Inference API
  • HuggingFace MCP Server: Built-in integration with Hugging Face's MCP server for accessing models, datasets, and spaces

Quick Start

  1. Run the agent:

    uvx hf-inference-acp
    
  2. If HF_TOKEN is not set, you'll start in Setup mode with these commands:

    • /login - Get instructions for HuggingFace authentication
    • /set-model <model> - Set the default model
    • /check - Verify your configuration
  3. Once authenticated (HF_TOKEN is set), you'll automatically start in Hugging Face mode.

  4. In Hugging Face mode, use /connect to connect to the Hugging Face MCP server for model/dataset search tools.

Curated Model Aliases

/set-model supports short aliases from fast-agent's curated Hugging Face list, including:

  • kimi
  • glm
  • minimax
  • deepseek4pro
  • kimi25 (thinking profile)
  • qwen35 (thinking profile)
  • qwen35instruct (instruct profile)

Kimi 2.5 aliases resolve to hf.moonshotai/Kimi-K2.5:fireworks-ai with curated defaults:

  • kimi25: temperature=1.0, top_p=0.95

Qwen 3.5 aliases resolve to hf.Qwen/Qwen3.5-397B-A17B:novita with curated sampling defaults:

  • qwen35: temperature=0.6, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0
  • qwen35instruct: temperature=0.7, top_p=0.8, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0

Configuration

Configuration is stored at ~/.config/hf-inference/hf.config.yaml:

default_model: hf.moonshotai/Kimi-K2-Instruct-0905

mcp:
  servers:
    huggingface:
      url: "https://huggingface.co/mcp?login"

Authentication

Set your HuggingFace token using one of these methods:

  1. Environment variable:

    export HF_TOKEN=your_token_here
    
  2. HuggingFace CLI:

    hf auth login
    

Get your token from: https://huggingface.co/settings/tokens

License

Apache License 2.0 - See the main repository for details.

More Information

For full documentation and the main project, visit: https://github.com/evalstate/fast-agent

Metadata

Release files for hf-inference-acp 0.10.48

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for hf-inference-acp 0.10.48
File Size Uploaded Provenance Actions
hf_inference_acp-0.10.48.tar.gz 29.7 kB Without provenance Details for hf_inference_acp-0.10.48.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for hf-inference-acp 0.10.48
File Interpreter ABI Platform Actions
hf_inference_acp-0.10.48-py3-none-any.whl Python 3 none any Details for hf_inference_acp-0.10.48-py3-none-any.whl

Total release size: 63.3 kB

Release files / hf_inference_acp-0.10.48.tar.gz

Download URL hf_inference_acp-0.10.48.tar.gz
Size 29.7 kB
Tags Source
SHA-256 checksum
How to use checksums
6e1215ae26a09dc30c2e0d8ffb8d4c644d2fef20ff0f31cd1b307833feb3da14
BLAKE2b-256 checksum
How to use checksums
b30eeb7db4fe696349fdeea8b24926f567cc757854d10b0d7472776d19fa7e83
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.13.5 {"installer":{"name":"uv","version":"0.13.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / hf_inference_acp-0.10.48-py3-none-any.whl

Download URL hf_inference_acp-0.10.48-py3-none-any.whl
Size 33.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
c072676fc31cf9f4deaeef830ee105b44b981526215e74f4887ffecb5c8bc3f1
BLAKE2b-256 checksum
How to use checksums
9d5f2531f7aaa632754dd6b0c2f455a4fcd46f120f3ca9047c60fc4078a62756
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.13.5 {"installer":{"name":"uv","version":"0.13.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Security for hf-inference-acp 0.10.48

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.10.48 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page