MCP Integration
Connect Claude, Cursor, and other AI clients to your Protobox tools via MCP
Every Protobox workspace exposes a managed MCP endpoint. Connect any MCP-compatible client — Claude Desktop, Cursor, Claude Code, or a custom agent — and it gets instant access to your tools, knowledge base, and prompts.
https://{workspace-slug}.protobox.app/mcpQuick Setup
Step 1: Generate the config
The CLI builds a ready-to-paste config from your stored API key and workspace:
protobox mcp config --format claude{
"mcpServers": {
"protobox": {
"url": "https://acme.protobox.app/mcp",
"headers": { "X-API-Key": "YOUR_API_KEY" },
"transport": "streamable-http"
}
}
}protobox mcp install generates the same snippet offline, without a server round-trip:
protobox mcp install --format claude--format | Target |
|---|---|
claude | Claude Desktop / Claude Code |
cursor | Cursor IDE |
vscode | Visual Studio Code (mcp install only) |
generic | Generic MCP client (default) |
Step 2: Add it to your client
Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"protobox": {
"type": "streamablehttp",
"url": "https://acme.protobox.app/mcp",
"headers": { "X-API-Key": "YOUR_API_KEY" }
}
}
}Restart Claude Desktop. Your Protobox tools appear in the tools menu.
Add to .cursor/mcp.json in your project:
{
"mcpServers": {
"protobox": {
"type": "streamablehttp",
"url": "https://acme.protobox.app/mcp",
"headers": { "X-API-Key": "YOUR_API_KEY" }
}
}
}Add to .mcp.json in your project root:
{
"mcpServers": {
"protobox": {
"type": "streamablehttp",
"url": "https://acme.protobox.app/mcp",
"headers": { "X-API-Key": "YOUR_API_KEY" }
}
}
}Replace acme with your workspace slug and YOUR_API_KEY with a key from that workspace.
Step 3: Test the connection
protobox mcp testThis runs a full protocol handshake: connect, list tools, list resources, and list prompts. Use protobox mcp info or protobox mcp status to see the server URL, resource counts, and health.
Scope with Toolsets
By default the endpoint exposes your workspace's enabled tools. For production, create a toolset so each client only sees the tools it needs.
# Create a toolset from a JSON definition
protobox toolsets create -f support-toolset.json
# Add tools to it
protobox toolsets add-tools <toolset-id> -t <tool-id-1>,<tool-id-2>
# Verify
protobox toolsets get <toolset-id>MCP Commands
| Command | Description |
|---|---|
protobox mcp | Combined info + status |
protobox mcp info | Server URL, status, and resource counts |
protobox mcp status | Health check the MCP server |
protobox mcp test | Run a full MCP protocol test |
protobox mcp config --format <client> | Output config for an AI client |
protobox mcp install --format <client> | Generate a config snippet offline |
protobox mcp tools <agent-id> | List MCP tools for an agent |
protobox mcp call <agent-id> <tool> [args] | Call an MCP tool by name |
protobox mcp inspect <agent-id> <tool> | Show a tool's schema |
Authentication
All MCP requests require an X-API-Key header. Keys are workspace-scoped.
| Header | Required | Value |
|---|---|---|
X-API-Key | Yes | Your workspace API key |
Content-Type | Yes | application/json |
Accept | Yes | application/json, text/event-stream |
Create keys in the dashboard under Connect → MCP or Manage → API Keys.
Security
Never commit API keys to version control. Use environment variables or your platform's secret management.
If a key is compromised, revoke it in the dashboard and generate a new one, then update your MCP config and restart clients.