Pydantic Skills plugin
Official plugins and skills from the Pydantic team: Logfire, Pydantic AI and Pydantic for Claude Code, Codex and Cursor
- 153 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The `logfire` plugin connects the hosted Logfire MCP server with your token and reads the project traces, logs and metrics.
- The `logfire-exporter` plugin sends Codex session data to Logfire: turns and tool calls with their content.
Install
- Works in
Manual install
claude plugin marketplace add pydantic/skillsThen `claude plugin install logfire@pydantic-skills` (or `ai`, `pydantic-ai-harness`, `pydantic`). For Logfire run `/logfire:connect`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The `logfire` plugin connects the hosted Logfire MCP server with your token and reads the project traces, logs and metrics.
- The `logfire-exporter` plugin sends Codex session data to Logfire: turns and tool calls with their content.
This is third-party code. Review the repository files before installing.
What it does
Pydantic Skills is a plugin marketplace from the Pydantic team itself. It holds five plugins. `logfire` adds observability: the agent places instrumentation in Python, JS/TS and Rust code, and the `/instrument`, `/debug` and `/query` commands in Claude Code work through traces and logs. `logfire-exporter` for Codex exports completed turns and tool calls to Logfire. The `ai`, `pydantic-ai-harness` and `pydantic` plugins give the agent knowledge of Pydantic AI, its Harness layer and the Pydantic library. The `ai` plugin also includes skills for migrating projects to Pydantic AI from Agno, Claude Agent SDK, Google ADK and LangChain, each with reference notes on concept mapping and verification before cutover. Separately, the `skills/` directory holds 12 standalone `SKILL.md` files following the agentskills.io standard, from `logfire-setup`, `logfire-query` and `logfire-evals` to `building-pydantic-ai-agents` and `pydantic`. They work in more than 30 agents, and released versions are published through Agent Skills Discovery at `pydantic.dev/.well-known/agent-skills`.
Who it is for. Python developers on Pydantic, Pydantic AI and Logfire who want the agent to write code the way the library authors recommend and to set up telemetry itself.
Good fit when
- Traces have to be added to a FastAPI or other framework app and placing Logfire calls by hand is tedious.
- An alert fired at night, and it is easier for the agent to work through traces with `/debug` than for you to scroll a dashboard.
- An agent built on LangChain or Google ADK is moving to Pydantic AI and the migration follows documented steps with a check before cutover.
- A data schema model is written in Pydantic and the agent needs current constraints and validators.
- Agent behavior in code should be tested against a dataset: that is what the `logfire-evals` skill is for.
Not a fit when
- Telemetry goes to a different system: the `logfire` plugins target Logfire and its MCP server.
- The project is not Python, JS/TS or Rust: Logfire instrumentation for other languages is not described.
- Session data must not leave the machine: the Codex exporter writes it to Logfire.
Example request
Add Logfire to the FastAPI service: instrument the app and database calls, then use /query to show the slowest traces of the last hour.Limitations
The `logfire` plugin requires a Logfire login: `/logfire:connect` in Claude Code, `logfire auth` or `LOGFIRE_TOKEN` in Cursor. The exporter needs a write token in the environment or in `~/.config/logfire-exporter/config.env`, after which Codex is restarted and the hooks are approved with `/hooks`. For the EU region in Codex the MCP address is switched to `https://logfire-eu.pydantic.dev/mcp`. The `ai`, `pydantic-ai-harness` and `pydantic` plugins are Claude Code only in the marketplace, and Cursor gets just `logfire`, installed by copying the folder.
How to disable. In Claude Code remove a plugin with `claude plugin uninstall <plugin>@pydantic-skills`. `/logfire:disconnect` stops session capture without removing the MCP server and skills. In Codex switch the plugin off in the plugin menu, in Cursor delete `~/.cursor/plugins/local/logfire`.
FAQ
How do I install the Pydantic plugins in Claude Code?
Add the marketplace with `claude plugin marketplace add pydantic/skills`, then install what you need: `claude plugin install logfire@pydantic-skills`, `ai@pydantic-skills`, `pydantic-ai-harness@pydantic-skills` or `pydantic@pydantic-skills`.
How is a plugin different from a file in the skills directory?
A plugin is tied to a specific agent: in Claude Code it brings the `/instrument`, `/debug` and `/query` commands, in Codex and Cursor skills and MCP tools. The `SKILL.md` files in `skills/` work in any agent that follows the agentskills.io standard.
What does logfire-exporter send?
Completed Codex turns and tool calls go to Logfire with a write token. If your sessions can contain secrets, decide first whether you are ready to store them there.
How do I connect the Logfire MCP in Codex for the EU region?
Remove the default entry with `codex mcp remove logfire`, add `codex mcp add logfire --url https://logfire-eu.pydantic.dev/mcp`, run `codex mcp login logfire` and start a new conversation so the tools reload.
Which migration skills are included?
The `ai` plugin holds skills for moving from Agno, Claude Agent SDK, Google ADK and LangChain to Pydantic AI, each with a reference on concept mapping and verification before cutover.
Discussion
No questions or reviews yet.
Related
PluginsA skills library for Claude Code, Cursor and Codex: team code review, TDD, stacked PRs and repository readiness scoring
Softaworks Agent Toolkit
Softaworks Agent Skills
A marketplace of skills, subagents and commands for coding agents: development, documentation, planning and workplace communication
15 skills for Claude Code and Codex: Laravel, Docker, UI, debugging, tests and docs in four ready plugin bundles
A set of independent skills against drift in large codebases: hard-cut, root-cause tracing, lean testing and CI right-sizing.
Claude Code Skills and Agents
Claude Code Skills & Agents
A Claude Code plugin bundling skills and sub-agents: deep research, image generation, GitHub work, code review and model switching
Five DSPy 3.2 skills for Claude Code and Codex: signatures, metrics, GEPA optimization, RLM and end-to-end pipeline assembly.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Pydantic Skills (this card) | no | 153 | |
| pydantic-deepagents | no | 1,117 | a ready-made CLI agent on Pydantic AI; here plugin and skill sets from the Pydantic team for your own agents |
| LangChain Skills | yes | 1,336 | skills for working with LangChain; here Pydantic, Pydantic AI and Logfire plus migration to Pydantic AI |
README in short
The README starts with a plugin table and three install sections: for Claude Code, Codex and Cursor. It then lists 12 standalone skills and explains how releases reach the Agent Skills Discovery index through CI.