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

# Configuration

> Reference the main arguments accepted by the LangChain SDK.

`groundforge_langchain.init()` accepts keyword arguments that define the Agent, GroundForge scope, connection, adapters, media handling, and service mode.

| Argument         | Required | Purpose                                                           |
| ---------------- | -------- | ----------------------------------------------------------------- |
| `agent`          | Yes      | LangChain or LangGraph Agent instance.                            |
| `endpoint`       | Yes      | GroundForge endpoint, normally `https://platform.groundforge.ai`. |
| `api_key`        | Yes      | Agent-targeted runtime API key.                                   |
| `agent_id`       | Yes      | Stable external Agent ID.                                         |
| `tenant_id`      | No       | Organization identifier.                                          |
| `workspace_id`   | No       | Workspace identifier.                                             |
| `name`           | No       | Friendly display name.                                            |
| `execution_mode` | No       | Supported execution mode for the Agent.                           |
| `params`         | No       | Runtime options and reported resources.                           |
| `logo_path`      | No       | PNG, JPEG, or WebP logo within the SDK size limit.                |
| `input_adapter`  | No       | Converts a Channel job to Agent input.                            |
| `output_adapter` | No       | Converts Agent output to the Channel response.                    |
| `media`          | No       | Media settings or custom handlers.                                |
| `serve`          | No       | Starts the Channel job connection.                                |

Common parameters include:

```python theme={null}
params={
    "channelInputMode": "auto",
    "approvalTimeoutSeconds": 300,
    "jobQueueSize": 100,
    "resources": [],
    "prompts": [],
}
```

Keep endpoint and identity values in environment configuration. Keep API keys and model credentials in a secret manager or protected environment variables — never in code.

Run one SDK initialization per Agent instance. Repeated initialization can create duplicate instrumentation or registration behaviour.

## Example

The demo repository shows the minimum connection values. See [LangChain Agent Example](/examples/langchain-agent) before adding optional adapters, media handlers, or service-mode settings.


## Related topics

- [Media](/sdk/media.md)
- [Usage Notes](/operations/usage-notes.md)
- [Weather MCP Example](/examples/weather-mcp.md)
- [Troubleshooting](/help/troubleshooting.md)