Skip to main content

Agent connector tutorial: Pydantic AI

In this tutorial, you'll create a new Python project with uv, add a Pydantic AI agent, equip it with one of Airbyte's agent connectors, and use natural language to explore your data. This tutorial uses GitHub, but if you don't have a GitHub account you can swap in any other agent connector and perform different operations.

Your agent executes through Airbyte. Airbyte Agents owns the OAuth apps, stores your third-party tokens, and refreshes them for you. Your Python code only ever sees your Airbyte client ID and client secret.

Overview​

This tutorial is for AI engineers and other technical users who work with data and AI tools. You can complete it in about 15 minutes.

The tutorial assumes you have basic knowledge of the following tools, but most software engineers shouldn't struggle with anything that follows.

  • Python and package management with uv
  • Pydantic AI
  • GitHub, or a different third-party service you want to connect to

Before you start​

Before you begin this tutorial, ensure you have the following.

  • Python version 3.13 or later
  • uv
  • An Airbyte Agents account. You can sign up for free.
  • Your Airbyte API credentials. Copy AIRBYTE_CLIENT_ID and AIRBYTE_CLIENT_SECRET from the Profile page in the Airbyte Agents web app. See Manage your user profile for details.
  • A GitHub connector added to your Airbyte Agents workspace. Add one of these two ways:
    • Web app (recommended): Go to Credentials in the Airbyte Agents web app, add a GitHub connector, and authenticate it with a GitHub personal access token (a classic token with repo scope is sufficient for this tutorial) or OAuth. See Add a connector for details.
    • API: Create a connector with POST /api/v1/integrations/connectors and store your GitHub credentials. See Add a connector for details.
  • An OpenAI API key. This tutorial uses OpenAI, but Pydantic AI supports other LLM providers if you prefer.

Part 1: Create a new Python project​

In this tutorial you initialize a basic Python project to work in. However, if you have an existing project you want to work with, feel free to use that instead.

Create a new project using uv:

uv init my-ai-agent --app
cd my-ai-agent

This creates a project with the following structure:

my-ai-agent/
├── .gitignore
├── .python-version
├── main.py
├── pyproject.toml
└── README.md

You create .env and uv.lock files in later steps, so don't worry about them yet.

Part 2: Install dependencies​

Install the Airbyte agent SDK, Pydantic AI, and python-dotenv:

uv add airbyte-agent-sdk pydantic-ai python-dotenv

This command installs:

  • airbyte-agent-sdk: The Airbyte Agents Python SDK, which ships every connector as a typed submodule.
  • pydantic-ai: The AI agent framework, which includes support for multiple LLM providers including OpenAI, Anthropic, and Google.
  • python-dotenv: A library you can use to load environment variables from a .env file.
note

If you want a smaller installation with only OpenAI support, you can use pydantic-ai-slim[openai] instead of pydantic-ai. See the Pydantic AI installation docs for more options.

Part 3: Import Pydantic AI and the GitHub agent connector​

  1. Create an agent.py file for your agent definition:

    touch agent.py

  2. Add the following imports to agent.py:

    agent.py
    from dotenv import load_dotenv
    from pydantic_ai import Agent
    from airbyte_agent_sdk import connect
    from airbyte_agent_sdk.connectors.github import GithubConnector

    These imports provide:

    • load_dotenv: Load environment variables from your .env file.
    • Agent: The Pydantic AI agent class that orchestrates LLM interactions and tool calls.
    • connect: The Airbyte agent SDK entry point. One call returns a typed connector bound to your workspace.
    • GithubConnector: The connector class. You reference it when decorating the tool so the SDK can describe the connector's entities and actions to the agent.

Part 4: Add a .env file with your secrets​

  1. Create a .env file in your project root and add your secrets to it. Replace the placeholder values with your actual credentials.

    .env
    AIRBYTE_CLIENT_ID=your-airbyte-client-id
    AIRBYTE_CLIENT_SECRET=your-airbyte-client-secret
    OPENAI_API_KEY=your-openai-api-key

    Copy AIRBYTE_CLIENT_ID and AIRBYTE_CLIENT_SECRET from the Profile page in the Airbyte Agents web app.

    warning

    Never commit your .env file to version control. If you do this by mistake, rotate your secrets immediately.

  2. Add the following line to agent.py after your imports to load the environment variables:

    agent.py
    load_dotenv()

    This makes your secrets available via os.environ. Pydantic AI automatically reads OPENAI_API_KEY from the environment, and the agent SDK picks up AIRBYTE_CLIENT_ID and AIRBYTE_CLIENT_SECRET from the environment in the next step.

Part 5: Configure your connector and agent​

Now that your environment is set up, add the following code to agent.py to create the GitHub connector and Pydantic AI agent.

Define the connector​

Connect to GitHub through your Airbyte Agents workspace:

agent.py
github = connect("github")

One line does four things for you:

  • Reads AIRBYTE_CLIENT_ID and AIRBYTE_CLIENT_SECRET from the environment.
  • Defaults to the "default" workspace, which is where the web app stores credentials unless you change it.
  • Returns a typed GithubConnector bound to the authenticated GitHub connector you added earlier.
  • Routes every github.execute(...) call through Airbyte's hosted API, which holds the GitHub OAuth tokens and refreshes them for you.

You never register an OAuth app, copy a GitHub token into your code, or write token-refresh logic.

If you want to connect to a different workspace or pass credentials explicitly, use connect("github", workspace_name="my-workspace", client_id=..., client_secret=...) or pass an AirbyteAuthConfig. See the SDK reference for details.

Define the agent​

Create a Pydantic AI agent with a system prompt that describes its purpose:

agent.py
agent = Agent(
    "openai:gpt-4o",
    system_prompt=(
        "You are a helpful assistant that can access GitHub data through the "
        "github_execute tool. Be concise and accurate."
    ),
)
  • The "openai:gpt-4o" string specifies the model to use. You can use a different model by changing the model string. For example, use "openai:gpt-4o-mini" to lower costs, or see the Pydantic AI models documentation for other providers like Anthropic or Google.
  • The system_prompt parameter is where you encode any API idiosyncrasies the model can't see in the tool schema. The Airbyte agent SDK already exposes entity names, actions, and enum values through the tool description, so the prompt only needs to carry domain constraints (pagination defaults, date formats, preferred streams) as your agent grows.
  • The prompt references a github_execute tool. You register that tool in the next part.

Part 6: Add a tool to your agent​

Rather than one tool per GitHub endpoint, the Airbyte agent SDK exposes the entire GitHub API through a single execute(entity, action, params) entry point. The tool_utils decorator fills in the entity and action catalog as the tool description, so the model knows what's available without you writing a schema.

Add the following to agent.py:

agent.py
@agent.tool_plain
@GithubConnector.tool_utils
async def github_execute(entity: str, action: str, params: dict | None = None):
    return await github.execute(entity, action, params or {})

The decorator stack is the whole tool definition. No per-action docstring, no GITHUB_LIST_COMMITS or GITHUB_GET_PR sprawl, one entry point that covers the full connector. @GithubConnector.tool_utils appends the full entity and action catalog to the tool description, and caps oversized responses. As the connector grows, the tool signature stays the same.

Each execute call returns a structured result with data (the records) and meta (pagination cursors). Pydantic AI serializes the dict for the model automatically, so you don't need to call json.dumps here. You can keep the result as-is, filter it in Python, or page through it using meta.end_cursor.

Part 7: Run your project​

Now that your agent is configured with a tool, update main.py and run your project.

  1. Update main.py. This code creates a simple chat interface in your command line tool and allows your agent to remember your conversation history between prompts.

    main.py
    import asyncio
    from agent import agent

    async def main():
        print("GitHub Agent Ready! Ask questions about GitHub repositories.")
        print("Type 'quit' to exit.\n")

        history = None

        while True:
            prompt = input("You: ")
            if prompt.lower() in ('quit', 'exit', 'q'):
                break
            result = await agent.run(prompt, message_history=history)
            history = result.all_messages()
            print(f"\nAgent: {result.output}\n")

    if __name__ == "__main__":
        asyncio.run(main())

  2. Run the project.

    uv run main.py

Chat with your agent​

The agent waits for your input. Once you prompt it, the agent decides which entity and action to call based on your question, asks Airbyte to execute it, and returns a natural language response. Try prompts like:

  • "List the 10 most recent open issues in airbytehq/airbyte"
  • "What are the 10 most recent pull requests that are still open in airbytehq/airbyte?"
  • "Are there any open issues that might be fixed by a pending PR?"

The agent has basic message history within each session, and you can ask followup questions based on its responses.

Troubleshooting​

If your agent fails to retrieve GitHub data, check the following:

  • HTTP 401/403 errors from Airbyte: Verify that AIRBYTE_CLIENT_ID and AIRBYTE_CLIENT_SECRET are copied correctly from your Profile page.
  • "No connector found" or "connector not configured": Make sure you've added a GitHub connector in the Credentials page of the Airbyte Agents web app. connect("github") defaults to the "default" workspace; if you added the connector to a different workspace, pass workspace_name="your-workspace-name" to connect().
  • HTTP 401/403 errors from GitHub: The GitHub token or OAuth credentials stored in your connector are invalid or missing required scopes. Open your GitHub connector in the web app and reauthenticate with a valid token that has repo scope.
  • Empty data=[] responses from filtered queries: Most GitHub filters use case-sensitive values. Confirm the agent is sending uppercase values (for example, states=["OPEN"] rather than states=["open"]). The system prompt in this tutorial nudges the model to do that by default.
  • OpenAI errors: Verify your OPENAI_API_KEY is valid, has available credits, and won't exceed rate limits.

See the Github agent connector page for more details.

Summary​

In this tutorial, you learned how to:

  • Set up a new Python project with uv
  • Add Pydantic AI and Airbyte's GitHub agent connector to your project
  • Configure environment variables for your Airbyte Agents credentials
  • Register a single tool that covers the entire GitHub API
  • Run your project and use natural language to interact with GitHub data through Airbyte

Next steps​

  • Add another connector. The same connect(...) + execute(...) pattern covers the full Airbyte agent connectors catalog. Add Slack, Stripe, Salesforce, or any other connector in the web app, then call slack = connect("slack") in your agent and register a second tool with another @agent.tool_plain / @SlackConnector.tool_utils stack. Your agent now reads GitHub and posts to Slack with no additional OAuth setup.
  • Use write actions. Connectors expose create, update, and post actions alongside the read ones. Ask the agent to file an issue, comment on a PR, or send a Slack message, and execute carries the write through with the stored OAuth token.
  • Let your AI assistant scaffold the next agent. The Airbyte agent SDK ships skills for Claude Code and Codex that carry the patterns above, so you can ask your assistant to build a new agent without retyping them. See the airbyte-agent-sdk repository for installation instructions.
  • Reach the same connectors from any MCP client. Airbyte Agents exposes the same connectors through a hosted MCP endpoint that works with Claude Code, Cursor, and ChatGPT. See the FastMCP tutorial for a local-server variant you can run yourself.
Was this page helpful?