It reads your app.
Your agent finds the backend, frontend, framework, local port, and existing auth routes. It decides the exact callback and session strategy.
Agent setup
No provider apps. No callback URLs. No signup per app. Give your coding agent the AnyOAuth skill and MCP server, then tell it which app to connect.
Add the AnyOAuth skill with:
npx skills add AnyOAuth/anyoauth --skill anyoauth
Connect the AnyOAuth MCP server with:
npx anyoauth-mcpPaste this into Claude Code, Codex, Cursor, or any agent that supports skills and MCP.
Create AnyOAuth login for example.com using AnyOAuth. Enable Google, GitHub and Magic Email Link.Replace example.com with your app or domain. Your agent inspects the app and does the rest.
Works with your tools
If it can run a skill and connect an MCP server, it can build the integration. No new editor, no separate dashboard workflow.
What happens next
Your agent handles the setup that used to live in three dashboards.
Your agent finds the backend, frontend, framework, local port, and existing auth routes. It decides the exact callback and session strategy.
It installs the skill and MCP server, creates or reuses one AnyOAuth project, registers the callback, and stores the client secret on the backend.
It adds the provider buttons, backend exchange, browser-bound PKCE, application session, and checks. You review the diff and sign in once.
Agent sign-in
AnyOAuth’s login services work for your agent as well as your users. Use Google or GitHub when a browser is available, or send a Magic Email Link to an inbox your agent can read.
Keep the session that started the login and open the link there. This is separate from the login code your agent writes for your app.
Less setup
AnyOAuth is built for agents: one skill, one MCP server, and one API for every application you own. Your projects stay in one account while the agent does the per-app work.
What agent-driven setup changes, and what stays in your control.
No. One AnyOAuth account covers unlimited applications. Your agent creates or reuses a project for each app, so there is no account-per-app and no repeated provider setup.
No. Your agent reads the app, determines the exact callback path, registers it with AnyOAuth, and writes the matching server route. You do not copy callback URLs between dashboards.
No. AnyOAuth manages the provider applications. Your agent creates the AnyOAuth project and uses the managed credentials; you never create a provider app ID or secret per project.
Any agent or IDE that can install a skill and connect an MCP server. The prompt works with Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode, VS Code, JetBrains, Zed, Neovim, Replit, v0, and other MCP clients.
It installs the AnyOAuth skill and MCP server, inspects the application, creates or reuses the project, registers callbacks, adds the backend and browser login code, stores the client secret on the backend, and runs the checks.
Use the same prompt but ask it to use the AnyOAuth CLI directly. The skill includes the CLI fallback and the generated command reference.
Not necessarily. Your agent can use Google or GitHub browser approval, a Magic Email Link sent to an inbox it can read (MailKite, SendHog, or AgentMailbox), or an owner-scoped API key for long-running automation.
Copy the prompt. Tell your agent which app. Review the result.