MCP Client Setup
Connect your favorite AI tools to clanker.engineer via MCP protocol.
Overview
The Model Context Protocol (MCP) allows AI assistants to use external tools. Once connected, you can:
- Search and install skills with natural language
- Install skills and monitor their runs (skills themselves are run through the REST API or in chat, not over MCP)
- Manage artifacts and executions
- Access your full clanker library
Prerequisites
- clanker account — Sign in with GitHub in the app. A card on file is required in every mode before a skill can run
- API key — Open Settings → External Agents and tap OR NAME A KEY (for static MCP configs), or tap ACTIVATE to pair a CLI agent that auto-receives a key via device-flow
Claude Desktop
Claude Desktop has native MCP support.
Configuration
-
Open Claude Desktop settings
-
Navigate to the MCP configuration file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
- macOS:
-
Add the clanker server:
{
"mcpServers": {
"clanker": {
"url": "https://next.clanker.engineer/mcp/",
"headers": {
"x-api-key": "YOUR_API_KEY"
}
}
}
}
- Restart Claude Desktop
Verification
Ask Claude:
“List my installed clanker skills”
Claude should call list-installed-skills and show your library.
Usage Examples
Search for skills:
“Search the clanker marketplace for code review skills”
Install a skill:
“Install the readme-generator skill from clanker”
Check on a run:
“Is my clanker README run finished? Show me the artifact.”
Check execution status:
“What’s the status of my current clanker execution?”
Cursor
Cursor IDE supports MCP for AI-assisted coding.
Configuration
- Open Cursor settings (Cmd/Ctrl + ,)
- Search for “MCP” or navigate to AI > MCP Servers
- Add a new server:
{
"clanker": {
"url": "https://next.clanker.engineer/mcp/",
"headers": {
"x-api-key": "YOUR_API_KEY"
}
}
}
- Save and reload Cursor
Usage in Cursor
When using Cursor’s AI features, you can reference clanker tools:
In chat:
“Use clanker’s code-reviewer skill to analyze this file for security issues”
With context:
“Find clanker’s test-generator skill and install it”
VS Code
VS Code can use MCP via extensions.
Setup with Continue Extension
- Install the Continue extension
- Open Continue settings
- Add MCP server configuration:
{
"mcpServers": {
"clanker": {
"url": "https://next.clanker.engineer/mcp/",
"headers": {
"x-api-key": "YOUR_API_KEY"
}
}
}
}
Setup with Other Extensions
Check your MCP-compatible extension’s documentation for configuration format. Most use similar JSON structure.
ChatGPT (with MCP Support)
ChatGPT supports MCP in Developer Mode and Deep Research.
Configuration
- Open ChatGPT settings
- Navigate to Developer Mode or Deep Research settings
- Add MCP server:
URL: https://next.clanker.engineer/mcp/
Headers:
x-api-key: YOUR_API_KEY
Custom Clients
Build your own MCP client using the protocol.
Streamable HTTP Transport (the only transport)
import { Client } from '@modelcontextprotocol/sdk/client';
import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp';
const transport = new StreamableHTTPClientTransport(
new URL('https://next.clanker.engineer/mcp/'),
{
headers: {
'x-api-key': process.env.CLANKER_API_KEY
}
}
);
const client = new Client({
name: 'my-client',
version: '1.0.0'
});
await client.connect(transport);
// List available tools (only those your key's scopes cover)
const tools = await client.listTools();
console.log('Available tools:', tools);
// Call a tool
const result = await client.callTool({
name: 'list-installed-skills',
arguments: {
limit: 5
}
});
console.log('Installed skills:', result);
A ck_ key may also be sent as Authorization: Bearer ck_….
From a CLANKER.NET dev container
A dev container provisioned by CLANKER.NET reaches Clanker over this same endpoint. Its key is scoped to what the container needs, and it is injected at the container’s egress proxy, so it never sits in the container’s environment or in your MCP config.
Clanker is the first product built on CLANKER.NET. Beyond developer containers, CLANKER.NET’s replicator lets a repository describe its own staging, production, static sites and storage in one file, run one stage at a time: a deploy shows its plan first, never deletes data unless told to, and protects production by default, with state and keys kept in your own account.
Available Tools
Once connected, you have access to all clanker MCP tools:
| Category | Tools |
|---|---|
| Skills | list-installed-skills, get-skill-details, install-skill, uninstall-skill, search-marketplace |
| Artifacts | list-artifacts, get-artifact, download-artifact, delete-artifact, save-artifact |
| Executions | get-execution-status, get-execution-diff, list-executions, cancel-execution |
| Fetch | web-fetch, http-fetch |
| Billing | get-credit-balance |
| Connectors | list-connectors, connect-connector, disconnect-connector, get-connector-status, search-connector-actions, run-connector-action, list-connector-triggers, list-sources, list-branches, get-source-content, list-github-repos |
| Workflows | start-workflow, search-workflows, list-schedules, set-schedule, unset-schedule, notification-inbox |
| Agents and traces | agent-step, list-traces, read-trace |
| Resources | install-resource, request-upload-url |
Each tool needs a scope; a tool your key’s scopes do not cover is hidden from
tools/list and refused on tools/call with an error naming the missing scope
(see Agent Permissions).
Skills are executed via REST (POST /api/v1/skills/:slug/run or
POST /api/v1/executions), not over MCP. Over MCP you discover, install, and monitor
runs; workflow runs are managed through the Workflows REST API.
See MCP API Reference for detailed documentation.
Troubleshooting
“Invalid API key” Error
- Verify the key in Settings → External Agents (it should appear with an ACTIVE / IDLE / DORMANT pill based on last-used time)
- Check for typos in configuration
- Ensure the key hasn’t been revoked
Connection Timeout
- Check your internet connection
- Verify the URL is correct (
https://next.clanker.engineer/mcp/)
Tools Not Appearing
- Restart your AI client
- Check configuration file syntax (valid JSON)
- Verify the server is listed in client’s MCP settings
Rate Limit Errors
MCP requests are limited to 60/minute. If you hit limits:
- Wait for the reset period
- Batch related operations
- Cache results when possible
Connection Drops
Long-lived connections may drop after ~30 minutes of inactivity. Your client should:
- Detect disconnection
- Automatically reconnect
- Resume from last known state
Security Best Practices
- Never share API keys - Each user should have their own key
- Use environment variables - Don’t hardcode keys in config files
- Rotate keys periodically - Delete and regenerate keys monthly
- Monitor usage - Check the app for unusual activity
- Revoke compromised keys - Immediately revoke if exposed
Next Steps
- Creating Skills - Build your own skills
- GitHub Actions CI/CD - Automate with CI/CD
- MCP API Reference - Explore all available tools