# Conduit developer integration

Conduit is an AI agent platform for hospitality operations. Use the REST API or Model Context Protocol server to give an authenticated application access to Conduit workspace data and actions.

## Installation

Conduit is a hosted service and does not require a local package. Create an API token under **Settings > API & MCP**, then choose REST or MCP for the client you are building.

## Configuration

REST requests use the base URL `https://api.conduit.ai/v1` and bearer authentication:

```bash
export CONDUIT_API_TOKEN="replace-with-your-workspace-token"
export CONDUIT_WORKSPACE_ID="replace-with-your-workspace-id"
```

For an MCP client, configure the Streamable HTTP endpoint `https://api.conduit.ai/v1/mcp`. Prefer read-only scopes unless the workflow genuinely needs to change data, and keep human approval for sensitive or irreversible actions.

## Usage

List the contacts available to the authenticated workspace:

```bash
curl --request GET \
  --url "https://api.conduit.ai/v1/contacts?workspace_id=$CONDUIT_WORKSPACE_ID" \
  --header "Authorization: Bearer $CONDUIT_API_TOKEN"
```

Use the OpenAPI document to generate typed clients and tool definitions rather than guessing request or response fields.

## Resources

- [Developer hub](https://www.conduit.ai/developers)
- [API documentation](https://docs.conduit.ai/api-reference/introduction)
- [OpenAPI specification](https://docs.conduit.ai/api-reference/openapi.json)
- [MCP documentation](https://docs.conduit.ai/mcp/overview)
- [LLM-readable index](https://www.conduit.ai/llms.txt)
- [Markdown sitemap](https://www.conduit.ai/sitemap.md)