Jentic MCP + Cursor¶
Goal — In a few minutes, you'll enable Jentic MCP tools in Cursor, allowing you to search, load, and execute external APIs directly in your coding workflow.
What you'll get¶
With Jentic MCP configured in Cursor, you'll gain access to powerful tools:
Tools available¶
| Tool available in Cursor | What it does |
|---|---|
| search_apis | Type what you need (e.g., "Send Discord Message", "Find New York Times Articles") and MCP lists matching APIs. |
| load_execution_info | See the details of any API you found - what information it needs and how to use it. |
| execute | Actually use the API - provide the required information and get results back in the chat. |
Before you start¶
Ensure you have the following:
| You need | Why | Quick install |
|---|---|---|
| Python ≥ 3.11 | Required for uv/uvx and Jentic CLI. | python.org |
| Cursor | The AI coding assistant where MCP tools run. | Cursor |
| uv / uvx runtime | Launcher that runs the Jentic MCP server. | Install with pip: pip install uv (any OS) or with Homebrew (macOS): brew install uvIf you have trouble with uvx, see Making Cursor See uvx below. |
1. Create an Agent & get your Agent API Key¶
- Go to app.jentic.com/sign-in and sign in to your account.
- In Credentials, add any API secrets you'll use (e.g., Discord bot token, NYT key). These credentials will be securely stored in Jentic and automatically used by your agent.
-
Open Agents → New Agent, select which APIs/workflows this agent can access, then Generate API Key.
-
Keys are scoped to the APIs you selected and shown once.
- Recommended: Store all API credentials in Jentic's Credentials section rather than using environment variables.
- Set the Agent API Key as an environment variable so Cursor can pass it to the MCP server:
macOS / Linux
export JENTIC_AGENT_API_KEY="<your-agent-api-key>"
Windows (PowerShell)
setx JENTIC_AGENT_API_KEY "<your-agent-api-key>"
$env:JENTIC_AGENT_API_KEY = "<your-agent-api-key>" # current session
✅ No more
jentic register,curl, orJENTIC_UUID. UseJENTIC_AGENT_API_KEYfrom the Agents page.
2. Configure Cursor to use Jentic MCP¶
-
Open Cursor.
-
Access your Cursor MCP configuration file (
mcp.json):The easiest way to open (or create if it doesn't exist) your
mcp.jsonis through the Cursor interface:- Via Cursor's Settings UI (Recommended Method):
- Click the cog wheel (⚙️) in the top right of the Cursor application.
- Select MCP from the sidebar.
- Click "Add new global MCP server".
- This will open (or create) your
mcp.jsonfile in your default text editor.
If you prefer, or if the UI method isn't available, you can manually find and open the file:
macOS/Linux:
~/.cursor/mcp.json- Terminal:
mkdir -p ~/.cursor open ~/.cursor/mcp.json - File Manager:
- Go to your home directory, show hidden files, open
.cursor/mcp.json(create if missing).
- Go to your home directory, show hidden files, open
Windows:
%USERPROFILE%\.cursor\mcp.json- File Explorer: paste
%USERPROFILE%\.cursor\in the address bar, createmcp.jsonif missing. - PowerShell:
notepad $env:USERPROFILE\.cursor\mcp.json
- Via Cursor's Settings UI (Recommended Method):
-
Add or update the Jentic MCP server configuration: Paste the following JSON into your
mcp.jsonfile. If the file already contains anmcpServersobject with other servers, carefully add thejenticentry within it, ensuring correct JSON formatting.
{
"mcpServers": {
"jentic": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/jentic/jentic-sdks.git@main#subdirectory=mcp",
"mcp"
],
"env": {
"JENTIC_AGENT_API_KEY": "<your-agent-api-key>"
}
}
}
}
-
Save the
mcp.jsonfile. -
Restart Cursor for the changes to take effect.
3. Managing API Credentials¶
Recommended approach: Store all API credentials in Jentic's Credentials section when creating your agent. This keeps credentials secure and centralized.
If you need to add credentials for APIs you want to use: 1. Go back to app.jentic.com and sign in. 2. Navigate to Credentials and add any additional API keys, tokens, or secrets needed. 3. Update your agent to include the new credentials, or create a new agent with the updated credentials.
Note: Only the
JENTIC_AGENT_API_KEYshould be set as an environment variable. All other API credentials should be stored in Jentic's Credentials section, not as environment variables in your MCP configuration.
Optional: Using Environment Variables¶
If you prefer to use environment variables directly instead of storing credentials in Jentic, you can add them to your MCP configuration:
- Open the same
mcp.jsonfile. - Insert or update the
envblock inside thejenticserver. - Save the file and restart Cursor.
Example configuration with environment variables:
{
"mcpServers": {
"jentic": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/jentic/jentic-sdks.git@main#subdirectory=mcp",
"mcp"
],
"env": {
"JENTIC_AGENT_API_KEY": "<your-agent-api-key>",
"DISCORD_BOTTOKEN": "<your-discord-bot-token>",
"OPENAI_API_KEY": "<your-openai-api-key>"
}
}
}
}
Security tip: The
mcp.jsonfile is plain text. Ensure it's stored securely, especially if it contains sensitive API keys.
4. If something doesn't work¶
| What you see / Problem | Try this |
|---|---|
| Jentic MCP tools not available in Cursor | Check mcp.json for JSON syntax errors (e.g., missing commas, incorrect brackets). Ensure Cursor was fully restarted after any changes. Check Cursor's logs for MCP-related errors. |
| “command not found: uvx” (in logs or error) | uvx is not in a path Cursor can see. See "Making Cursor See uvx" below for details. |
| API calls fail (e.g., 401/403 errors) | The API likely requires authentication. Use /load_execution_info to check its requirements, then add the necessary credentials to Jentic's Credentials section and ensure your agent has access to them. Restart Cursor. |
| Other errors from Jentic MCP tools | Check Cursor's logs for more details. For advanced debugging, you can run the Jentic MCP server manually and inspect its direct output/logs. |
Making Cursor See uvx¶
Cursor, like many applications, might not automatically pick up binaries added to your shell's PATH via files like .zshrc or .bashrc. Choose one of these options:
Option A – Homebrew / System package (macOS & Linux)¶
- Homebrew (macOS) or your distro’s package manager / a
sudo pip install uvdropsuvxsomewhere Cursor already looks:/usr/local/binon Intel Macs & most Linux distros/opt/homebrew/binon Apple Silicon
If which uvx returns a path, you’re good – skip this section.
Option B – pip / pipx / pyenv (user-local installs)¶
If uvx lives in a user-local directory, expose it via a symlink (macOS/Linux) or by adding the folder to PATH (Windows).
- macOS / Linux – symlink
sudo ln -sf "$(which uvx)" /usr/local/bin/uvx # use /opt/homebrew/bin on Apple Silicon sudo ln -sf "$(which uv)" /usr/local/bin/uv - Windows – PATH
- Locate the folder containing
uvx.exe(e.g.%USERPROFILE%\AppData\Roaming\Python\Python311\Scripts). - Add that folder to Path in Environment Variables….
- Restart Cursor (and any open terminals).
- Locate the folder containing
Option C – Hard-code the full path¶
Skip PATH tweaks entirely and set the absolute path to uvx / uvx.exe in the command field of mcp.json.