Value Proposition Provide AI agents with reliable, real-time visibility into connection pools and route health. Seamlessly monitor InnoDB clusters. Optimize database traffic routing. This ensures high availability and resilient agent operations. Read the full value proposition.

• A recent version of MySQL Router with REST API enabled.
• Router REST API credentials (username/password)
• Network access to Router REST API endpoint (default: HTTPS on port 8443)
• For InnoDB Cluster mode: The cluster must be running for REST API authentication

Includes, but is not limited to:

⚠️ = Requires InnoDB Cluster. Note: The mysql_router_pool_status tool does not require an InnoDB Cluster. It only relies on the router's active connection pool configuration.

Bootstrapping Router against an InnoDB Cluster uses metadata_cache authentication. This authenticates API users against the metadata database.

1. Create REST API user in the cluster:

-- Connect to any cluster node
-- The user will be stored in mysql_innodb_cluster_metadata.router_rest_accounts

# Use mysqlrouter_passwd to generate the password hash
mysqlrouter_passwd set /tmp/router_admin.pwd router_admin
# Manually read the generated hash from /tmp/router_admin.pwd and replace <hash_from_mysqlrouter_passwd> below
# Insert into cluster metadata (on PRIMARY node)
mysql -h localhost -P 3307 -u cluster_admin -p -e "INSERT INTO mysql_innodb_cluster_metadata.router_rest_accounts(cluster_id, user, authentication_method, authentication_string, description) SELECT cluster_id, 'router_admin', 'modular_crypt_format', '<hash_from_mysqlrouter_passwd>', 'REST API user' FROM mysql_innodb_cluster_metadata.clusters LIMIT 1;"

2. Router config uses metadata_cache backend:

[http_auth_backend:default_auth_backend]
backend=metadata_cache

For standalone Router deployments without InnoDB Cluster:

[http_server]
port=8443
ssl=1
ssl_cert=/path/to/router-cert.pem
ssl_key=/path/to/router-key.pem

[http_auth_realm:default_auth_realm]
backend=default_auth_backend
method=basic
name=default_realm

[http_auth_backend:default_auth_backend]
backend=file
filename=/path/to/mysqlrouter.pwd

[rest_router]
require_realm=default_auth_realm

[rest_routing]
require_realm=default_auth_realm

# Generate password hash (prompts for password)
mysqlrouter_passwd set /path/to/mysqlrouter.pwd router_admin

Secure MySQL Router operations using OAuth validation. See the OAuth page for setup instructions.

{
  "mcpServers": {
    "mysql-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@neverinfamous/mysql-mcp",
        "--transport",
        "stdio",
        "--mysql",
        "mysql://user:password@localhost:3306/database"
      ],
      "env": {
        "MYSQL_ROUTER_URL": "https://router.example.com:8443",
        "MYSQL_ROUTER_USER": "router_admin",
        "MYSQL_ROUTER_PASSWORD": "router_password",
        "MYSQL_ROUTER_INSECURE": "true"
      }
    }
  }
}

If you only want Router tools (e.g., for a dedicated monitoring agent):

{
  "args": [
    "--transport",
    "stdio",
    "--mysql",
    "mysql://user:password@localhost:3306/database",
    "--tool-filter",
    "router"
  ]
}

Cause: Router REST API is unreachable or TLS handshake failed.

Solutions:

1. Verify Router is running: docker ps | grep router
2. Check Router logs: docker logs mysql-router
3. Test API manually: curl -k -u router_admin:router_password https://localhost:8443/api/<version>/router/status
4. Ensure MYSQL_ROUTER_INSECURE=true is set for self-signed certificates

Cause: Authentication failed. This usually means the InnoDB Cluster is not running.

Solutions:

1. Start the InnoDB Cluster: docker compose -f test-server/infrastructure/innodb-cluster.yml up -d
2. Reboot cluster from outage if needed: dba.rebootClusterFromCompleteOutage()
3. Restart Router to reconnect: docker restart mysql-router
4. Verify credentials match the router_rest_accounts table

Cause: Cluster nodes are running but Group Replication is not active.

Solution: Reboot cluster from complete outage using MySQL Shell:

mysqlsh --uri cluster_admin:password@localhost:3307 --js \
  -e "dba.rebootClusterFromCompleteOutage('clusterName', {force: true})"

Cause: The connection pool name doesn't exist or connection pooling is disabled.

• Tools - Complete tool list
• Tool Filtering - Custom tool filtering
• Configuration - General configuration