Integration Guide
MCP endpoints, authentication methods, and connection examples for integrating AI agents with MCP Gateway.
MCP Endpoint URL
Each MCP gets a unique endpoint URL displayed on its card. Point your AI agent to this URL to connect.
https://mcp.quilr.ai/mcp/<your-mcp-slug>/
Authentication Methods
MCP Gateway supports three authentication methods depending on the MCP server's requirements.
Token-Based Authentication
The most common method. Send a Bearer token with a user identifier header for per-user tracking.
Authorization: Bearer <token>
mcpuser: user@company.com
- Create API tokens in Settings → API Tokens
- Each token is scoped to a specific agent
- The
mcpuserheader identifies the end user for per-user tracking
OAuth - Dynamic Client Registration (DCR)
The gateway automatically registers as an OAuth client with the MCP server. No Client ID or Secret needed - just click Connect and authorize.
OAuth - Manual
For MCP servers without DCR support. Provide your Client ID and Client Secret during MCP setup. The gateway uses these credentials for the authorization flow.
Token-Based Connection Example
# Connect to MCP endpoint with token auth
curl -X POST https://mcp.quilr.ai/mcp/your-mcp-slug/ \
-H "Content-Type: application/json" \
-H "Authorization: Bearer <your-api-token>" \
-H "mcpuser: user@company.com" \
-d '{
"jsonrpc": "2.0",
"method": "tools/list",
"id": 1
}'
Agent Configuration
- Each AI agent (OpenAI, Claude, Cursor, etc.) connects to the MCP endpoint URL shown on the MCP card
- The gateway identifies agents by their User-Agent header keyword
- Configure which MCPs each agent can access via Agents Configuration or per-MCP Access Control