# fast-agent

> Code, Build and Evaluate agents - excellent Model and Skills/MCP/ACP Support

- **URL**: https://www.freshcrate.ai/projects/fast-agent
- **Author**: evalstate
- **Category**: MCP Servers
- **Latest version**: `v0.7.15` (2026-06-02)
- **License**: Apache-2.0
- **Source**: https://github.com/evalstate/fast-agent
- **Homepage**: https://fast-agent.ai
- **Language**: Python
- **GitHub**: 3,750 stars, 402 forks
- **Registry**: github
- **Tags**: `acp`, `agent`, `agent-framework`, `agent-skills`, `cli`, `mcp`, `mcp-client`, `mcp-server`, `python`

## Description

Code, Build and Evaluate agents - excellent Model and Skills/MCP/ACP Support

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v0.7.15` | 2026-06-02 | High | ## What's Changed  * SEP-2640, "Skills over MCP" by @olaservo (thank you!) https://github.com/evalstate/fast-agent/pull/815 * attach_media rendering in tool loop   **Full Changelog**: https://github.com/evalstate/fast-agent/compare/v0.7.14...v0.7.15 |
| `v0.7.13` | 2026-05-28 | High | ## What's Changed  🚨Security Update - starlette pinned to v1.0.1 🚨  * Anthropic Opus 4.8 * Install skills direct from path/repo url * TUI doc (asciinema flow v1) by @evalstate in https://github.com/evalstate/fast-agent/pull/807 * Render path tidy ups * Fix/imports by @evalstate in https://github.com/evalstate/fast-agent/pull/808 * Repair docs and integration tests by @evalstate in https://github.com/evalstate/fast-agent/pull/812 * Fix security report #811   **Full Changelog**: http |
| `v0.7.11` | 2026-05-24 | High | ## What's Changed  - OSC133 improvements - Terminal Render image attachments (in/out) - `--attach` for the CLI - Update TUI documentation (first step of longer term goals) - fix: Responses handling of tool input types - Deepseek updates -- `deepseek` now uses the Deepseek first-party provider, and model updates for deepseek v4.  - Reduce prompt-toolkit default gutter size (8 to 6)  **Full Changelog**: https://github.com/evalstate/fast-agent/compare/v0.7.7...v0.7.11 |
| `v0.7.7` | 2026-05-19 | High | ## Gemini Flash 3.5  Support for Gemini Flash 3.5, plus web_search (google_search) on native Google endpoints.  ## What's Changed * model, core concepts, styling updates, zensical fix by @evalstate in https://github.com/evalstate/fast-agent/pull/797 * test environment failures, docs tweaks by @evalstate in https://github.com/evalstate/fast-agent/pull/798 * gemini flash 3.5, file resolution improvements for CLI by @evalstate in https://github.com/evalstate/fast-agent/pull/799   **Full C |
| `v0.7.5` | 2026-05-17 | High | ## What's Changed  * Batch Processing support for parallel execution and Hugging Face datasets (including Parquet). Install with `fast-agent-mcp[parquet]` * Separate Plugins from Card Packs, allow global install through `$FAST_AGENT_HOME` `fast-agent.yaml`. REQUIRES v0.7.5 and above. * Refreshed Documentation Site - migrate to zensical, place in `docs` folder (`fast-agent-docs` is now retired). * Improve Tool Listing  * Ty update by @evalstate in https://github.com/evalstate/fast-agent/pul |
| `v0.7.2` | 2026-05-10 | High | ## HF Bucket / URL Support  `hf://` URIs are now supported for most CLI options and System/Prompt templating.   From CLI:  `fast-agent --card hf://buckets/evalstate/demo-bucket/ai-news-summary-card.md`  From a template:  ```markdown  You are a Helpful AI agent.  Take note of the below content: ---  {{url:hf://<url>}}  --- ```  ## Model Updates  - xAI Grok models support provider supplied `x_search` tool (use `grok?x_search=true`), with WebSockets enabled by default.  - ` |
| `v0.7.0` | 2026-05-05 | High | ## fast-agent 0.7.0  - Added `fast-agent batch structured` for JSONL/CSV batch jobs with JSON Schema or Pydantic schema output, resumable   runs, sampling, per-row errors, telemetry, and summaries. - Added `DeepSeek4` and `MiniMax 2.7` for Hugging Face Inference Providers - Expanded structured-output support across providers, including JSON Schema loading, Pydantic schema models,  `structured_tools` -policies, and a `fast-agent check structured-tools` compatibility probe. - Use viterbi tun |
| `v0.6.25` | 2026-04-27 | High | ## Privacy Filter Export (powered by openai/privacy-filter)  Use the new OpenAI Privacy Filter model to sanitize agent trace exports, and upload directly to Hugging Face. Install with  `uv tool install fast-agent-mcp[privacy]`  Download the privacy filter model with:  `fast-agent export --privacy-filter --download-privacy-filter`  Export, filter and upload your latest trace with: `fast-agent export --privacy-filter --hf-dataset <username>/<repo>`  ## GPT-5.5 via API Support  **Ful |
| `v0.6.22` | 2026-04-21 | High | ## HF Trace Export  Export session traces in a Hugging Face Trace Viewer compatible format - and directly upload to a dataset. Available from `/session export` or `fast-agent export --hf-dataset <repo>`.   ## Kimi 2.6 and Kimi 2.6 Instant  Model aliases added for above models via Novita.   ## Other Changes  - Improvements to ACP session, error and system prompt handling.  **Full Changelog**: https://github.com/evalstate/fast-agent/compare/v0.6.21...v0.6.22 |
| `v0.6.17` | 2026-04-16 | High | ## What's Changed  ### Opus 4.7 Support  Includes "summarized display" feature to show reasoning summaries.  - Adaptive reasoning is on by default. Set Reasoning Level and Task Budget with `/model` commands or - via model string - e.g. `opus?reasoning=xhigh&task_budget=64000`.  Access earlier versions with presets `opus46` and `opus45` or the full model slug.   Note that task_budget is only available for Opus 4.7  ### Assistant Response Banner  Long responses now show a banner at t |

## Citation

- HTML: https://www.freshcrate.ai/projects/fast-agent
- Markdown: https://www.freshcrate.ai/projects/fast-agent.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/fast-agent/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._