Use an AI assistant for supported Opslin deployment work.

The beta MCP server exposes supported Opslin operations such as deployment status, logs, application health, domains, databases, alerts, backup and recovery workflows. Availability depends on the connected account and the capability being used.

A connection is approved through Opslin before an AI client receives access. MCP calls are evaluated against the same organisation and permission boundaries used by Opslin, rather than receiving unrestricted server access.

Connection instructions are provided to eligible beta users inside Opslin. Public client-specific setup instructions and an official MCP Registry listing will be published only after the public endpoint and supported-client matrix are verified.

The public beta documents a Claude Code connection flow. Other MCP-compatible assistants may be evaluated when they support OAuth 2.1 with PKCE and Streamable HTTP, but Opslin does not publish a universal client-compatibility promise yet.

Read actions can inspect app status, deployment details, logs, health, metrics, domains, databases, backups, alerts, and firewall previews. Write actions can deploy, roll back, start or stop supported resources, manage domains, backups, alerts, databases, and firewall changes when the connected token has the required scope.

Sensitive or destructive actions remain permission-controlled. Database password reveal and destructive database operations are not treated as general read access.

Examples include: ‘Show the latest deployment for my app’, ‘Why did the last release fail?’, and ‘Check whether my app is healthy’. Eligible users receive the MCP server URL and the Claude Code setup command inside the dashboard, then approve the OAuth connection before the client receives access.

The current MCP server identifies as Opslin version 1.0.0 in the protocol handshake. A public MCP changelog and registry listing are deferred until the endpoint, client matrix, and approval behavior have completed public verification.