pydantic-ai / openai
AGENTS.md21k starsChanged 2 days agoMITView on GitHub
These rules add to models/AGENTS.md and apply to the OpenAI package only.
Adopted by
Steady: 0 new projects took it in the last 4 weeks, none the 4 weeks before.thin data
Who's talking about pydantic/pydantic-ai
- A Hacker News story“MCP Run Python”Hacker News · Apr 15, 2025 · 173 points
- @pythonhub.dev“PydanticAI PydanticAI is a Python Agent Framework designed to make it less painful to build production …”Bluesky · Dec 11, 2024 · 8 likes
- langfuse/langfuse“Add support for `gen_ai.system_instructions` gen-ai field”GitHub · Sep 19, 2025 · 7 reactions
- A Hacker News story“Agent Framework / shim to use Pydantic with LLMs”Hacker News · Dec 2, 2024 · 7 points
- pydantic/mcp-run-python“Dockerise the MCP Run Python server”GitHub · May 27, 2025 · 5 reactions
The biggest 5 of 64 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 64
Talked about on GitHub, Bluesky, Hacker News and the web (64 times).
On GitHub · 50
- Sep 19, 2025Add support for `gen_ai.system_instructions` gen-ai field — langfuse/langfuse7 reactions · 4 comments
- May 27, 2025Dockerise the MCP Run Python server — pydantic/mcp-run-python5 reactions · 6 comments
- Oct 16, 2025Inngest Durable execution with Pydantic-AI — inngest/inngest-py4 reactions · 9 comments
- May 23, 2025Handle `type[T]` for type variables — astral-sh/ty4 reactions · 3 comments
- Oct 11, 2026Pydantic AI integration: support TemporalDurability (TemporalAgent is deprecated, removed in v3) — temporal-community/temporal-agent-harness3 reactions · 1 comment
45 more on GitHub
- Mar 12, 2026Impossible to run the app due to Mistral import error — Denario-AI/Denario3 reactions · 3 comments
- Feb 27, 2026Anthropic cache tokens double-counted: usage.input already includes cache via OTel/genai-prices — langfuse/langfuse3 reactions · 6 comments
- May 25, 2025Misc. bug: Streaming with tools causes pydantic-ai to mess up tool name — ggml-org/llama.cpp3 reactions · 5 comments
- May 21, 2026chore(deps): bump pydantic-ai from 1.56.0 to 1.99.0 — dorel14/SoniqueBay2 reactions · 1 comment
- May 21, 2026chore(deps): bump pydantic-ai from 1.56.0 to 1.99.0 in /backend/api — dorel14/SoniqueBay2 reactions · 1 comment
- Apr 1, 2026chore(deps): bump pydantic-ai from 0.2.15 to 1.56.0 in /backend — cicero-im/meeting-minutes2 reactions · 1 comment
- Apr 1, 2026Bump pydantic-ai-slim from 1.0.1 to 1.56.0 — cicero-im/codantic-aidisplayagent2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai from 1.0.1 to 1.56.0 — cicero-im/codantic-aidisplayagent2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai-slim from 0.0.46 to 1.56.0 in /pydantic-ai-langfuse — cicero-im/ottomator-agents2 reactions · 1 comment
- Apr 1, 2026Bump pydantic-ai-slim from 0.6.2 to 1.56.0 — cicero-im/paigeant2 reactions · 2 comments
- Apr 1, 2026chore(deps): bump pydantic-ai from 0.2.6 to 1.56.0 — cicero-im/tunacode2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai-slim from 0.7.4 to 1.56.0 in /agent-backend — cicero-im/pydantic-ai-fastapi-react-vite-agent2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai from 1.0.10 to 1.56.0 in /examples/integrations/pydanticAI_researcher_example — cicero-im/memori2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai from 1.0.17 to 1.56.0 — cicero-im/github-issue-analysis2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai from 0.6.2 to 1.56.0 — cicero-im/paigeant2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai from 0.0.46 to 1.56.0 in /pydantic-ai-langfuse — cicero-im/ottomator-agents2 reactions · 1 comment
- Apr 1, 2026chore(deps): bump pydantic-ai-slim from 1.0.15 to 1.56.0 — cicero-im/pydantic_ai_claude_code2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai from 0.7.4 to 1.56.0 in /agent-backend — cicero-im/pydantic-ai-fastapi-react-vite-agent2 reactions · 1 comment
- Apr 1, 2026Bump pydantic-ai-slim from 0.0.30 to 1.56.0 — cicero-im/pydanticai_mcp_demo2 reactions · 1 comment
- Apr 1, 2026build(deps): bump pydantic-ai from 0.2.0 to 1.56.0 — cicero-im/lightblue-ai2 reactions · 1 comment
- Apr 1, 2026Bump pydantic-ai from 0.3.1 to 1.56.0 — cicero-im/pydantic_ai_editor2 reactions · 2 comments
- Apr 1, 2026chore(deps): bump pydantic-ai from 0.3.4 to 1.56.0 — cicero-im/mcp2 reactions · 2 comments
- Apr 1, 2026chore(deps): bump pydantic-ai-slim from 0.3.4 to 1.56.0 — cicero-im/mcp2 reactions · 3 comments
- Apr 1, 2026chore(deps): bump pydantic-ai-slim from 0.2.12 to 1.56.0 — cicero-im/pydantic-ai-mcp-experiment2 reactions · 1 comment
- Apr 1, 2026build(deps): bump pydantic-ai-slim from 0.2.18 to 1.56.0 — cicero-im/me-pydanticai-2 reactions · 1 comment
- Apr 1, 2026Bump pydantic-ai from 0.0.30 to 1.56.0 — cicero-im/pydanticai_mcp_demo2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai-slim from 0.3.1 to 1.56.0 — cicero-im/pydantic_ai_editor2 reactions · 2 comments
- Apr 1, 2026build(deps): bump pydantic-ai from 0.2.18 to 1.56.0 — cicero-im/me-pydanticai-2 reactions · 2 comments
- Apr 1, 2026build(deps): bump pydantic-ai-slim from 0.0.52 to 1.56.0 — cicero-im/wa_llm2 reactions · 1 comment
- Apr 1, 2026build(deps): bump pydantic-ai-slim from 0.2.0 to 1.56.0 — cicero-im/lightblue-ai2 reactions · 1 comment
- Apr 1, 2026Bump pydantic-ai-slim from 0.0.43 to 1.56.0 — cicero-im/Ai_Tutor2 reactions · 1 comment
- Apr 1, 2026chore(deps): bump pydantic-ai from 0.0.42 to 1.56.0 — cicero-im/automagik-agents2 reactions · 2 comments
- Apr 1, 2026Bump pydantic-ai from 0.0.43 to 1.56.0 — cicero-im/Ai_Tutor2 reactions · 2 comments
- Apr 1, 2026build(deps): bump pydantic-ai from 0.0.52 to 1.56.0 — cicero-im/wa_llm2 reactions · 1 comment
- Mar 26, 2026build(deps): bump the pip group across 5 directories with 4 updates — GoldenEagle-23/oss-fuzz2 reactions · 1 comment
- Feb 28, 2026Error because `thoughtSignature` missing from Gemini 3 function call — Arjun-Vashistha/Test-Repo-22 reactions
- Nov 21, 2025Run generator.__anext__ in the same task to avoid anyio RuntimeError — langfuse/langfuse-python2 reactions · 1 comment
- Nov 14, 2025Dockerise MCP-run-python server + add git workflows to publish images — pydantic/mcp-run-python2 reactions · 5 comments
- Aug 4, 2025Tracking: Contribute FastMCP Toolset to Pydantic AI — PrefectHQ/fastmcp2 reactions · 1 comment
- Jul 12, 2025Tool context kwarg detection not compatible with string-valued/__future__ annotations — modelcontextprotocol/python-sdk2 reactions · 3 comments
- May 23, 2025Sequence of types — python/mypy2 reactions · 2 comments
- Nov 6, 2026API proxy token usage reports 0 input and cache tokens for an Anthropic-compatible Z.AI endpoint — github/gh-aw1 reaction
- May 29, 2026Consider automation to close duplicate PRs? — jupyterlab/jupyterlab1 reaction · 1 comment
- May 7, 2026chore: port pydantic-ai v1.91.0 changes — a7ul/vibes1 reaction
- May 5, 2026chore: port pydantic-ai v1.90.0 changes — a7ul/vibes1 reaction
On Bluesky · 7
- Dec 11, 2024PydanticAI PydanticAI is a Python Agent Framework designed to make it less painful to build production grade applications with Generative… — @pythonhub.dev8 likes and reposts
- Aug 9, 2025I use 'llm' now for my cli explorations with genai. Will i move to `clai` for that "FastAPI feeling" they promise? i do love pydantic! see:… — @paulkarayan.bsky.social2 likes and reposts
- Dec 4, 2024🚀 Skyrocketing! 🚀 (200+ new stars) 📦 pydantic / pydantic-ai ⭐ 1,319 (+429) 🗒 Python Agent Framework / shim to use Pydantic with LLMs — @github-trending.bsky.social2 likes and reposts · 1 reply
- Sep 3, 2025https://github.com/pydantic/pydantic-ai — @tom-doerr.bsky.social1 likes and repost
- Jul 18, 2025📦 pydantic / pydantic-ai ⭐ 10,912 (+49) 🗒 Python Agent Framework / shim to use Pydantic with LLMs — @github-trending.bsky.social1 likes and repost
2 more on Bluesky
- Dec 6, 2024PydanticAI: Build production-grade Agentic AI apps in pure Python! PydanticAI offers the same elegance and ease of use as FastAPI, now… — @akshay-pachaar.bsky.social1 likes and repost · 1 reply
- Dec 4, 2024github.com/pydantic/pyd... — @presidentlin.bsky.social0 likes and reposts · 1 reply
On Hacker News · 5
- Apr 15, 2025MCP Run Python173 points · 66 comments
- Dec 2, 2024Agent Framework / shim to use Pydantic with LLMs7 points · 3 comments
- Dec 3, 2024Pydantic AI3 points
- Dec 6, 2024PydanticAI using Ollama (llama3.2) running locally2 points · 3 comments
- Dec 5, 2024PydanticAI1 point
On the web · 2
- Nov 11, 2026pydantic-ai/README.md at main — nl.mitryy.ru
- Oct 30, 2026Pydantic AI - Review & Setup Guide - Developers Digest — developersdigest.tech
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository48 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- docs/harness/AGENTS.mdAGENTS.md
- docs/model_aliases/AGENTS.mdAGENTS.md
- docs/realtime/AGENTS.mdAGENTS.md
- .github/workflows/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/capabilities/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/durable_exec/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/models/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/models/anthropic/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/native_tools/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/profiles/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/providers/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/realtime/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/toolsets/AGENTS.mdAGENTS.md
- pydantic_ai_slim/pydantic_ai/ui/AGENTS.mdAGENTS.md
- src/pydantic_ai_harness/AGENTS.mdAGENTS.md
- src/pydantic_clai2/AGENTS.mdAGENTS.md
- tests/AGENTS.mdAGENTS.md
- tests/benchmarks/AGENTS.mdAGENTS.md
- tests/harness/AGENTS.mdAGENTS.md
- adding-a-provider-api-featureSkill
- add-new-modelSkill
- adopt-prSkill
- branch-contextSkill
- complete-partial-prSkill
- docs-parity-reviewerSkill
- file-followup-issueSkill
- i-have-adhdSkill
- initialize-worktreeSkill
- poweruser-feature-auditSkill
- pre-push-reviewSkill
- pushing-commits-to-the-repoSkill
- refresh-issue-briefSkill
- testing-skillSkill
- building-pydantic-ai-agentsSkill
- migrating-agno-to-pydantic-aiSkill
- migrating-claude-agent-sdk-to-pydantic-aiSkill
- migrating-google-adk-to-pydantic-aiSkill
- migrating-langchain-to-pydantic-aiSkill
- migrating-mastra-to-pydantic-aiSkill
- migrating-openai-agents-sdk-to-pydantic-aiSkill
- migrating-pi-to-pydantic-aiSkill
- migrating-vercel-ai-sdk-and-eve-to-pydantic-aiSkill
- migrating-deep-agents-to-pydantic-ai-harnessSkill
- pydantic-ai-harnessSkill
- reviewerSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 0 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 6
Never import from __init__.py at runtime in a submodule.
- Very specific: line 7
Keep the runtime imports between submodules acyclic.
- Very specific: line 8
Import each submodule that defines a public API name inside the try block of __init__.py that imports the openai SDK.
- Very specific: line 9
Import _batch.py and _responses_websocket.py at runtime only inside the function that uses them, never at a module's top level.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
pydantic_ai_slim/pydantic_ai/models/openai/ Guidelines
These rules add to models/AGENTS.md and apply to the OpenAI package only.
- Keep each model class and each streamed response in its own submodule:
OpenAIChatModelin_chat_model.py,OpenAIStreamedResponsein_chat_streaming.py,OpenAIResponsesModelin_responses_model.py,OpenAIResponsesStreamedResponsein_responses_streaming.py. Put a module-level helper in the_*.pysubmodule whose docstring names its concern. Do not split a class's methods into mixins: a mixin spreads one object's state across files. - Never import from
__init__.pyat runtime in a submodule.__init__.pyimports the submodules, so that import is circular. - Keep the runtime imports between submodules acyclic. When a submodule needs a name only in annotations and a runtime import would close a cycle, import the name under
TYPE_CHECKING, as_chat_model.pydoes forOpenAIResponsesModel. - Import each submodule that defines a public API name inside the
tryblock of__init__.pythat imports theopenaiSDK. A missing SDK then raises the install hint, not a bareImportError. - Import
_batch.pyand_responses_websocket.pyat runtime only inside the function that uses them, never at a module's top level._responses_websocket.pyneeds theopenai-realtimeextra, and an error importing_batch.pymust not read as a missingopenaiSDK. - In a test, patch a name in the module that calls it:
pydantic_ai.models.openai._chat_model.download_itemfor a call in_chat_model.py. Eachfrom ... import namecreates its own binding, so a patch onpydantic_ai.models.openaireaches no call.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

