README.md•2.29 kB
# CB Insights MCP Server
The CBI MCP Server provides an interface for developers to interact with CB Insights ChatCBI LLM through AI Agents.
## Tools
### ChatCBI
- Sends a message from an agent to our AI chatbot and returns the response
- Input parameters:
- `message`:
- `chatID`: (optional) The unique id of an existing ChatCBI session. Used for continuity in a conversation. If not provided, a new ChatCBI session will be created
- Returns object containing the following fields:
- `chatID`: Unique id of current ChatCBI session
- `message`: ChatCBI message generated in response to the message send in the input.
- `RelatedContent`: Content that is related to the content returned
- `Sources`: Supporting sources for the message content returned
- `Suggestions` Suggested prompts to further explore the subject matter
- For more information, check the [ChatCBI Docs](https://api-docs.cbinsights.com/portal/docs/api#tag/ChatCBI)
## Setup
The CBI MCP Server uses [uv](https://docs.astral.sh/uv/getting-started/installation/) to manage the project.
The default port is `8000`, but can be modified by updating the `CBI_MCP_PORT` environment variable in the `.env` file.
The timeout for requests can also be modified via the `CBI_MCP_TIMEOUT` variable in the `.env` file.
### Authentication
Documentation on how CB Insights APIs are authenticated can be found [here](https://api-docs.cbinsights.com/portal/docs/CBI-API/Authentication)
The server uses the `CBI_CLIENT_ID` and `CBI_CLIENT_SECRET` environment variables set in the `.env` file to authorize requests.
## Usage
### With Claude Desktop
Update the `claude_desktop_config.json` file using the following command:
```shell
mcp install server.py
```
This will add the following configuration:
```json
{
"mcpServers": {
"cbi-mcp-server": {
"command": "/path/to/.local/bin/uv",
"args": [
"--directory",
"/path/to/cloned/cbi-mcp-server",
"run",
"server.py"
]
}
}
}
```
## Debugging
The [inspector](https://modelcontextprotocol.io/docs/tools/inspector#getting-started) can be used to test/debug your server.
```shell
mcp dev server.py
```
[More info on using the inspector](https://modelcontextprotocol.io/docs/tools/inspector#py-pi-package)