GitHubDashboard
Docs/MCP Gateway

MCP Gateway

Squidward acts as a governed MCP gateway. Configure your AI tools to route all MCP calls through Squidward for policy enforcement, DLP, and audit logging.

How It Works

AI Client

Cursor, Claude, etc.

Authenticates with Squidward token

Squidward Gateway

Policy evaluation
DLP scanning
Credential injection
Audit logging

Provider MCP

GitHub, Vercel, etc.

Receives governed requests

Gateway Features

Policy Enforcement

Every MCP call is evaluated against your security policies in real-time.

Credential Security

Provider credentials never reach AI clients. Squidward injects them securely.

Full Audit Trail

Every request and response is logged for compliance and debugging.

DLP Scanning

Requests and responses are scanned for sensitive data patterns.

Gateway Endpoint

Configure your AI tools to use this MCP gateway URL:

https://gateway.squidward.app/mcp
bash

All requests must include an Authorization header with your Squidward token:

Authorization: Bearer mcp_live_xxxxxxxxxxxxx
http

Configuration Examples

Configure your AI tool to use Squidward as the gateway:

// ~/.cursor/mcp.json
{
  "mcpServers": {
    "squidward": {
      "url": "https://gateway.squidward.app/mcp",
      "headers": {
        "Authorization": "Bearer ${SQUIDWARD_TOKEN}"
      }
    }
  }
}
json

Generate Configuration

Use the CLI to automatically generate and install configurations:

# Generate and install for Cursor
squidward config generate cursor

# Generate and install for Claude
squidward config generate claude

# Generate for any tool (prints to stdout)
squidward config generate --format json
bash

Automatic Token Handling

When you generate a config, the CLI uses environment variable references ($${SQUIDWARD_TOKEN}) by default. Set this in your shell profile or the tool's environment.

Environment Setup

Set your Squidward token as an environment variable:

# Add to ~/.zshrc or ~/.bashrc
export SQUIDWARD_TOKEN=mcp_live_xxxxxxxxxxxxx

# Reload shell
source ~/.zshrc
bash

Verify Connection

Test that your configuration is working:

# Test via CLI
squidward mcp list

# Output:
# Connected to Squidward Gateway
# Available MCP Tools:
# ├── github:list_repos
# ├── github:create_issue
# ├── vercel:list_projects
# └── ...
bash

Then test from your AI tool by invoking any MCP. The response will include governance metadata confirming it went through Squidward.

Tool-Specific Guides

Troubleshooting

Connection refused / timeout

Ensure your network allows outbound HTTPS connections to gateway.squidward.app. Check firewall rules and proxy settings.

401 Unauthorized

Your token is invalid or expired. Generate a new token with squidward tokens createor re-authenticate with squidward login.

403 Policy Denied

The request was blocked by a security policy. Use squidward policies simulateto understand why and adjust policies if needed.