> ## Documentation Index
> Fetch the complete documentation index at: https://docs.civic.com/llms.txt
> Use this file to discover all available pages before exploring further.

# smolagents

> Connect a smolagents agent to Civic's MCP Hub using streamable HTTP transport

Connect a [smolagents](https://huggingface.co/docs/smolagents/) agent from Hugging Face to Civic using `MCPClient` with Streamable HTTP transport. smolagents' MCP client discovers all available tools and exposes them to a `CodeAgent`.

## Prerequisites

* Python 3.11+
* A Civic account at [app.civic.com](https://app.civic.com) with a configured toolkit
* A Civic token and an Anthropic API key

## Installation

```bash theme={null}
pip install "smolagents[mcp]" litellm python-dotenv
```

## Environment Variables

```bash theme={null}
CIVIC_URL=https://app.civic.com/hub/mcp?profile=your-toolkit
CIVIC_TOKEN=your-civic-token
ANTHROPIC_API_KEY=your-anthropic-key
```

<Card title="Get Your Credentials" icon="key" href="/civic/quickstart/credentials">
  How to generate a Civic token and configure toolkit URL parameters
</Card>

## Connecting to Civic

Pass a single config dict to `MCPClient` with the transport key set to `"streamable-http"` (hyphen, not underscore):

```python theme={null}
import os
from dotenv import load_dotenv
from smolagents import CodeAgent, LiteLLMModel
from smolagents.mcp_client import MCPClient

load_dotenv()

mcp_client = MCPClient(
    {
        "transport": "streamable-http",  # hyphen required — not streamable_http
        "url": os.environ["CIVIC_URL"],
        "headers": {"Authorization": f"Bearer {os.environ['CIVIC_TOKEN']}"},
    }
)
tools = mcp_client.get_tools()
print(f"{len(tools)} tools loaded")

model = LiteLLMModel(model_id="anthropic/claude-sonnet-4-6")
agent = CodeAgent(tools=tools, model=model)
```

<Note>
  The transport key must be `"streamable-http"` with a hyphen. Using `"streamable_http"` with an underscore will fail. Pass a flat dict to `MCPClient` — not a nested `{"server_name": {...}}` structure.
</Note>

## Running the Agent

```python theme={null}
result = agent.run("What events do I have today?")
print(result)

# Always disconnect when done
mcp_client.disconnect()
```

## Production Configuration

For production agents, lock to a specific toolkit using the `profile` URL parameter:

```bash theme={null}
CIVIC_URL=https://app.civic.com/hub/mcp?profile=your-production-toolkit
```

## Reference Implementation

<Card title="smolagents-reference-implementation-civic" icon="github" href="https://github.com/civicteam/smolagents-reference-implementation-civic">
  Complete implementation with FastAPI chat UI and deployment guide
</Card>

## Next Steps

<CardGroup cols={2}>
  <Card title="Agent Deployment" icon="robot" href="/civic/quickstart/clients/agents">
    Production deployment guide: profile locking, URL params, authentication
  </Card>

  <Card title="Guardrails" icon="shield" href="/civic/concepts/guardrails">
    Constrain what tools your agent can call
  </Card>

  <Card title="Audit Trail" icon="list-check" href="/civic/concepts/audit">
    Query what your agent did via Civic Chat
  </Card>

  <Card title="Get Credentials" icon="key" href="/civic/quickstart/credentials">
    Token generation and URL parameter reference
  </Card>
</CardGroup>