v0.4.22

Agent Cards

New feature - Agent Cards.

Read the article by @iqdoctor here: https://github.com/evalstate/fast-agent/blob/main/plan/agentcard-standards-mini-article.md.

Agents can be loaded as peers, or as tools (using --card / --card-tool or /card <filename> [--tool] from ACP/UI. Cards are also auto-loaded from .fast-agent/agent-cards or .fast-agent/tool-cards, and can reference Python function tools.

• REPL. New AgentCard in md fromat and CLI loading/lazy hot swap with --agent-cards | --watch | --reload options. New workflow-md samples by @iqdoctor in #585
• Add command line option for loading card as tool by @evalstate in #590
• Feat/auto load by @evalstate in #593
• Add decorator function_tools support and RAG example by @iqdoctor in #592

Hugging Face Space Deployment

Improved support for HF Spaces deployment (deployment skill to be follow) - use HF OAuth to use end-user token for inference.

• Feat/hf outh by @evalstate in #595

What's Changed

• Fix/markdown stream improvements, uvloop by @evalstate in #582
• prompt toolkit hide user message by @evalstate in #583
• fix model resolution by @evalstate in #584
• Respect skills=None as explicit disable by @iqdoctor in #588
• change test running for speed by @evalstate in #587
• Feat/markdown experiment1 by @evalstate in #591

Full Changelog: v0.4.17...v0.4.22

v0.4.17

Sampling with Tools Release

• SEP-1577 - Sampling with Tools. @evalstate (#578)
• SEP-1036 - URL Elicitaitions.
• SEP-991 - CIMD support for OAuth.

Notes

You may need to remove earlier DCR tokens before re-authorizing. Use fast-agent auth to manage saved tokens.

v0.4.16

What's Changed

• Feat/skills manager by @evalstate in #567
• Feat/token display acp by @evalstate in #568
• Fix/types by @evalstate in #570
• fix diffs for non-streaming models by @evalstate in #571
• simplify streaming arguments by @evalstate in #572
• refactor to tool runner loop by @evalstate in #562
• minimax 2.1 and GLM 4.7 by @evalstate in #573

Full Changelog: v0.4.13...v0.4.16

v0.4.13

fast-agent 0.4.13 - What's Changed

MAKER Agent Type by @lucidprogrammer

MAKER (“Massively decomposed Agentic processes with K-voting Error Reduction”) wraps a worker agent and samples it repeatedly until a response achieves a k-vote margin over all alternatives (“first-to-ahead-by-k” voting). This is useful for long chains of simple steps where rare errors would otherwise compound.

https://fast-agent.ai/agents/workflows/#maker

https://arxiv.org/abs/2511.09030

Agents-as-Tools by @iqdoctor

https://fast-agent.ai/agents/workflows/#agents-as-tools

The Agents As Tools workflow takes a complex task, breaks it into subtasks, and calls other agents as tools based on the main agent instruction.
This pattern is inspired by the OpenAI Agents SDK Agents as tools feature.

Agent Skills Installer / Refresh to Standard

The /skills command now lets you add, remove and install skills from valid registries. Hugging Face and Anthropic are supplied by default, and can be overriden in the configuration file. Seen here with Toad. Some small tweaks to the existing skills implementation to bring it in line with https://agentskills.io/home.

Other Changes

Also thanks to @iqdoctor for collecting community resources here: https://fast-agent.ai/ref/fast_agent_ai_articles/. Let us know if there are more to add 🔥.

• fix gpt-4o document modalities
• add gemini 3 flash support
• other improvements for hf-inference-acp (model selection etc.)

Detailed Change List

• Review ACP implementation with new SDK Union types by @evalstate in #549
• Improve Google provider: Vertex AI (ADC) support & preview model fallback by @iqdoctor in #550
• Make agents ACP context aware by @evalstate in #551
• feat: add @fast.maker decorator for reliable multi-step workflows by @lucidprogrammer in #548
• Feat: hybrid Agents-as-Tools/MCP-as-tools experimental agent by @iqdoctor in #515
• fix(google): default maxTokens per Gemini model by @iqdoctor in #554
• Feat/hf acp by @evalstate in #555
• docs(examples): align agents_as_tools_simple with docs by @iqdoctor in #553
• Fix missing Hugging Face agent without token by @evalstate in #557
• update mcp library,gpt-4o pdf fix by @evalstate in #558
• Add model list display to set-model command by @evalstate in #556
• Add inference provider lookup for models by @evalstate in #559
• fix(hf-inference-acp): always set model and show provider info by @evalstate in #561
• Review MCP Server connection in agents ACP by @evalstate in #563
• Include full tool args in ACP titles by @iqdoctor in #565
• Implement agent skills standard support by @evalstate in #564
• gemini 3 flash by @evalstate in #566

Full Changelog: v0.4.4...v0.4.13

v0.4.4

What's Changed

• Add custom refinement instruction on @fast.evaluator_optimizer by @bandinopla in #538
• feat: Add video support for Google Gemini provider by @lucidprogrammer in #537
• Feat/acp sdk update by @evalstate in #543
• MCP SEP-1330: Elicitation schema updates for Enums by @chughtapan in #324
• OpenAI Providers custom HTTP Headers by @evalstate in #544

New Contributors

• @bandinopla made their first contribution in #538
• @lucidprogrammer made their first contribution in #537

Full Changelog: v0.4.3...v0.4.4

v0.4.3

What's Changed

General

• Model selection available through FAST_AGENT_MODEL environment variable, message on startup when used.
• Enhanced Tool Call / Transport instrumentation in timing channel
• Allow absolute paths to be specified for Skills. (NB - best results usually when skills are in a subdirectory).
• LLM Retry Logic (thank you @usamaJ17 !)
• MCP Session Termination retry (use reconnect_on_disconnect: False in config to disable)
• Enhanced support/testing for HF Reasoning models.
• Improve HF provider selection (and provider selection message on startup).

Agent Client Protocol

• Auth view/reset
• Improved Streaming and Permission updates.
• Improved Reasoning token streaming.

Auto-generated:

• fix: prompt template file resolution ordering by @evalstate in #521
• update tool timing saving, including transport channel by @evalstate in #523
• allow absolute paths for skills directories by @evalstate in #524
• Feat/model env option by @evalstate in #526
• Feat/reasoning streaming by @evalstate in #529
• Implement Agent Client Protocol tools for CLI by @evalstate in #528
• fix: Centralize robust API retry logic in FastAgentLLM & preserve context by @usamaJ17 in #517
• Dev/0.4.2 by @evalstate in #530
• Compare session termination handling implementations by @evalstate in #532
• Feat/auth status acp by @evalstate in #533
• Improve acp progress by @evalstate in #534
• Simplify status command when no errors exist by @evalstate in #535

New Contributors

• @usamaJ17 made their first contribution in #517

Full Changelog: v0.4.1...v0.4.3

v0.4.1

What's Changed

• Fix #518. Remove ESC key handling - ctrl+c still cleanly cancels generation.
• Opus 4.5 support (alias opus and opus45)

v0.4.0

fast-agent 0.4.0

Agent Client Protocol (ACP) Support

fast-agent now has support for Zed Industries Agent Client Protocol, allowing seamless integrations with editors like Zed and Toad.

Read more about the integration at https://fast-agent.ai/acp/.

Workflow agents and MCP Tool Calls deliver workflow and progress notifications, multimodal content, agent skills, model comparison and cancellation are all supported. The default system prompt loads AGENTS.md if present.

A new command line fast-agent-acp has been added for convenience, and a no-install package is on PyPi - use it with uvx fast-agent-acp@latest --model xxx.

Stateless LLM Providers

LLM Providers are now stateless, simplifying cross-model switching and history management. The load_history command now has completions to enable easy loading.

Other Changes

• ESC key cancellation is supported
• Hugging Face kimi kimithink and qwen3 aliases now use Together as their default provider.

What's Changed

• Stateless LLM Providers by @evalstate in #499
• Review and audit codebase by @evalstate in #503
• Refactor type hints to modern Python syntax by @evalstate in #502
• Add ESC key cancellation for LLM provider calls by @evalstate in #505
• Remove unnecessary getattr calls with type info by @evalstate in #506
• Review ACP cancellation token implementation by @evalstate in #507
• Use prompt toolkit completion for /load_history command by @evalstate in #509
• Add ESC key handler to cancel streaming messages by @evalstate in #510
• Add load slash command to Python SDK by @evalstate in #511
• Add telemetry to Chain and Iterative Planner agents by @evalstate in #512
• Dev/0.4.0 by @evalstate in #514

Full Changelog: v0.3.29...v0.4.0

v0.3.29

What's Changed

• gpt-5.1 support
• add kimithink alias for Hugging Face, clamp kimi and kimithink aliases to "together" provider.
• Improve System Prompt template handling https://fast-agent.ai/agents/instructions/

Review full changelog here:

Full Changelog: v0.3.23...v0.3.29

v0.3.23

Hugging Face Inference Providers Support 🤗

HF Inference Providers are now fully supported -- read the documentation here: https://fast-agent.ai/models/llm_providers/#hugging-face

NB The default alias for kimi now points to the hf provider, with kimigroq for the Groq direct provider.

The following aliases have been configured, and tested/configured with Structured Outputs and Tool Use:

Alias	Maps to
kimi	hf.moonshotai/Kimi-K2-Instruct-0905
gpt-oss	hf.openai/gpt-oss-120b
gpt-oss-20b	hf.openai/gpt-oss-20b
glm	hf.zai-org/GLM-4.6
qwen3	hf.Qwen/Qwen3-Next-80B-A3B-Instruct
deepseek31	hf.deepseek-ai/DeepSeek-V3.1
minimax	hf.MiniMaxAI/MiniMax-M2

To try simply use fast-agent --model kimi for HF Routing or fast-agent --model kimi:together to specify "Together AI" as the provider.

Package Changes

Bedrock and Tensorzero packages should now be installed as add-ons:

uv tool install fast-agent-mcp[bedrock] and uv tool install fast-agent-mcp[tensorzero] respecitvely.

Other Changes

• ConversationSummary is a new utility class that analyzes agent conversation history and provides computed statistics for easy monitoring, debugging and evaluations. It automatically extracts key metrics including tool call counts, error rates, per-tool breakdowns, and timing information (average response time, total elapsed time, conversation span). All statistics are accessible as properties. The class can be used with the load_history function to analyse saved conversations.
• Change Server-Tool separator from - to __.
• Improvements to History display
• save_history now uses a default filename if not specified.

What's Changed

• Feat/serve command by @evalstate in #445
• fix: remove unused tomli package by @danfimov in #448
• fix issue #447 by @evalstate in #449
• fix(common): Improve constant SEP from "-" to "__" for namespaced names by @codeboyzhou in #298
• Feat/tool naming display by @evalstate in #451
• Fix recent changes defect by @evalstate in #452
• Test agent job execution and performance metrics by @evalstate in #453
• fix: Prevent display freeze during tool call streaming by @evalstate in #454
• fix: Restore buffer truncation to enable continuous scrolling by @evalstate in #455
• docs: rename plantype references to plan_type by @iqdoctor in #456
• Add FAST_AGENT_TIMING channel for instrumentation by @evalstate in #457
• Reduce package dependencies in fast-agent by @evalstate in #459
• Fix save_history function bug by @evalstate in #460
• Feat/acp by @evalstate in #463
• Fix All Remaining Issues by @evalstate in #462
• Add ACP terminal support to agent implementation by @evalstate in #464
• Handle overlapping ACP prompt requests gracefully by @evalstate in #465
• Fix duplicate messages in ACP client connection by @evalstate in #466

New Contributors

• @danfimov made their first contribution in #448
• @iqdoctor made their first contribution in #456

Full Changelog: v0.3.17...v0.3.23