Explorium AgentSource MCP Server
by explorium-ai
# Explorium API MCP Server
This MCP server is used to interact with the Explorium API.
**Note:** this is the README for developing the MCP server. For usage instructions, see the [README-pypi.md](README-pypi.md).
## Setup
Clone the repository:
```bash
git clone https://github.com/explorium-ai/mcp-explorium.git
cd mcp-explorium
```
Install uv and activate the virtual environment:
```bash
pip install uv
uv sync --group dev
```
## Running Locally
When developing locally, use `local_dev_server.py` to expose the MCP server to local clients.
You can test it with the MCP Inspector:
```bash
mcp dev local_dev_server.py
```
Create an `.env` file in the root of the repository with an
`EXPLORIUM_API_KEY` environment variable if it's not already configured.
### Usage with Claude Desktop
Follow the official guide to install Claude Desktop and set it up to use MCP servers:
https://modelcontextprotocol.io/quickstart/user
Then, add this entry to your `claude_desktop_config.json` file:
```json
{
"mcpServers": {
"Explorium": {
"command": "<UV_INSTALL_PATH>",
"args": [
"run",
"--directory",
"<REPOSITORY_PATH>",
"mcp",
"run",
"local_dev_server.py"
],
"env": {
"EXPLORIUM_API_KEY": "<YOUR_API_KEY>"
}
}
}
}
```
Be sure to replace all the `<PLACEHOLDERS>` with the actual values.
Run `which uv` to get your `uv` install path.
### Usage with Cursor
Cursor has [built-in support for MCP servers](https://docs.cursor.com/context/model-context-protocol).
To configure it to use the Explorium MCP server, go to
`Cursor > Settings > Cursor Settings > MCP` and add an "Explorium" entry
with this command:
```bash
uv run --directory repo_path mcp run local_dev_server.py
```
Make sure to replace `repo_path` with the actual path to the repository.
You may turn on "Yolo mode" in Cursor settings to use tools without
confirming under `Cursor > Settings > Cursor Settings > Features > Chat > Enable Yolo mode`.
**Note:** local usage will not work if dev dependencies are not installed,
or if the virtual environment is not activated.
## Building and Deploying
To build the MCP server, bump the version in `pyproject.toml`, then run:
```bash
uv build
```
This will create a `dist` directory with the built package.
You may then deploy the package to PyPI using the following command:
```bash
twine upload dist/*
```