MCP Server — Connect Any Coding Agent or Chat Client

VibeHost ships a remote Model Context Protocol server at https://api.vibehost.com/mcp. Any MCP-compatible client can connect and gain a tool surface for the entire platform.

Where MCP fits

VibeHost has two integration channels. Pick based on what's driving the work:

• Coding agents sit in your editor / terminal. They get the smoothest experience: native MCP support plus the ability to shell out to the vibehost CLI directly when scripted control beats tool calls. Ten first-class integrations, each with its own guide:
  • Claude Code — Anthropic, terminal-first
  • Codex CLI — OpenAI, terminal-first
  • Cursor — IDE, most-adopted
  • Antigravity — Google, agentic IDE
  • Antigravity CLI — Google, open-source terminal agent
  • Windsurf — Cognition (Devin's makers), agentic IDE
  • OpenCode — SST, open-source terminal agent
  • GitHub Copilot CLI — GitHub, terminal
  • Hermes Agent — Nous Research, self-improving terminal agent
  • Grok Build — xAI, terminal-first
• Chat clients (ChatGPT, Claude Desktop) don't have a CLI channel. They reach VibeHost only through MCP. Setup is one URL — see Chat clients below.

The rest of this page covers the MCP server itself: tool list, OAuth, per-client setup, error codes. If you're picking a coding agent, the per-agent guides above are a shorter path.

What your agent can do

A tool per major action, split across 8 surfaces. Read-only tools are marked (read) below. Tools that change publicly-visible internet state carry openWorldHint: true so the client can prompt for confirmation.

Choosing an upload path

A deploy needs your files on the server. There are two ways to get them there, and the right one depends on whether your client's environment can make outbound HTTPS requests directly to api.vibehost.com — not on whether it has a local filesystem.

The normal path uploads file bytes by PUTting them straight to api.vibehost.com (via short-lived signed URLs), outside the MCP channel, then deploys a manifest. That's faster and dedupes content across deploys — use it whenever you can reach the API:

1. Hash your files and call request_upload to mint signed upload URLs.
2. PUT each blob to its URL.
3. Deploy with the manifest.

But some clients can't make that direct PUT — a sandbox with no network egress, or a chat client like ChatGPT whose only channel to VibeHost is the MCP tool call itself. For those, write the files server-side with the draft-tree tools, then deploy with no manifest:

• create_file(appId, path, content) — write or overwrite a whole file (POSIX-relative path, e.g. index.html; content capped at 1 MB/file).
• edit_file(appId, path, oldString, newString, replaceAll?) — exact-string replacement (fails if oldString is absent, or occurs more than once without replaceAll).
• read_file / list_files — inspect the draft tree.
• delete_file — remove a file from it.

Then deploy with no manifest — the server assembles it from the draft tree. The draft persists after deploy, so you can edit_file and re-deploy incrementally. Writing a draft requires the same deployer grant as deploying.

The decision is network reachability, not filesystem: an agent that has files on disk but no egress to api.vibehost.com still belongs on the draft path.

Endpoints

Transport: Streamable HTTP (the MCP spec's recommended transport since 2025-06).

Authentication

OAuth 2.1 authorization-code flow with mandatory PKCE (S256). The first time your agent connects:

1. Agent calls /.well-known/oauth-protected-resource/mcp to discover the authz server.
2. Agent registers itself dynamically at /api/v1/oauth/register (no client secret needed for public clients).
3. Agent kicks off the auth code flow with PKCE.
4. You approve in browser; agent gets an access token.
5. Access token is 60 minutes, refresh token is 30 days with single-use rotation and reuse detection.

Token audience is bound to https://api.vibehost.com/mcp via RFC 8707, so a token minted for VibeHost won't be accepted by another resource server.

Coding agents (recommended)

These have native MCP, run in your editor / terminal, and can ALSO call the vibehost CLI directly when scripted control beats tool-calling. Each has a dedicated guide with the recommended setup.

Chat clients (MCP-only)

These don't have a CLI channel — MCP is the only way to drive VibeHost from them. Setup is one URL each.

Other MCP-compatible clients

Continue, Cline, Roo Code, Devin, Zed, Aider, Amazon Q, VS Code (Copilot Chat), JetBrains AI Assistant, Raycast, Warp, OpenClaw (chat-app gateway), and anything else that speaks MCP — point it at https://api.vibehost.com/mcp. Native custom-scheme redirects (claude-desktop:, cursor:, vscode:, windsurf:, raycast:, etc.) are pre-allowlisted, so OAuth lands cleanly without extra config.

Permissions

Tool calls run through the same per-app permissions the dashboard uses. The OAuth token represents you; the agent can only do what you can do.

• viewer grant → read-only tools work
• deployer grant → also deploy, promote, rollback
• admin grant → also manage settings + grants

Tools that change public state (create_deployment, add_custom_domain, set_password_gate, etc.) carry openWorldHint: true so well-behaved clients prompt you before calling.

Privacy

The MCP server adds no extra data store. Every tool call:

• Authenticates against the same user/token records as the dashboard.
• Audit-logs the same way (operation, actor, resource, outcome).
• Hashes OAuth refresh tokens at rest.

Tools that touch third-party services (add_grant invites via Resend; verify_custom_domain uses public DNS resolvers) declare openWorldHint: true.

See VibeHost privacy — same policy applies.

Tools → CLI cheatsheet

Every MCP tool maps 1:1 to a CLI command. Useful for "what would this prompt actually do?":

Example prompts

Once connected, you can talk to your agent in natural language. The agent picks the right tool(s).

For destructive operations (delete_app, delete_channel, delete_deployment), well-behaved clients prompt you to confirm before calling. The destructiveHint: true annotation is part of the tool definition.

Confirmation prompts — what gets asked

Tools annotated openWorldHint: true change publicly-visible state and should trigger a confirmation in your MCP client. Whether they do depends on the client:

• Claude Desktop / Claude Code — always prompts for destructive + open-world tools. Configurable per-session ("always allow").
• Cursor — prompts on first use; remembers per-tool.
• ChatGPT connectors — prompts on first call; user can mark as "trusted".
• Codex / programmatic clients — depends on host; some skip confirmation by design (assume agent autonomy is OK).

If you're scripting against MCP and don't want a human in the loop, use a PAT against the REST API instead. Tools are just convenience wrappers around the same endpoints.

Common errors

See also

• For a programmatic alternative without OAuth, see Personal access tokens — a PAT drives the same REST surface over HTTP.