Skip to content
AppHandoff

Hosted MCP Server

Hosted MCP server with live project context for AI coding agents

AppHandoff's MCP server exposes over 30 tools that give AI agents structured access to API contracts, database schemas, scan results, and handoff tickets — no file scraping, no stale documentation.

What is AppHandoff's MCP server?

An MCP server is a service that exposes tools and resources to AI coding agents through the Model Context Protocol — the open standard originally developed by Anthropic. AppHandoff's hosted MCP server connects directly to your project repositories and exposes pre-analyzed context: endpoint contracts, database schemas, frontend-backend mismatches, handoff tickets, and deployment status.

Instead of asking an agent to infer project state from raw files or Slack threads, the AppHandoff MCP server delivers machine-readable answers to the questions agents actually ask: what endpoints exist, which ones are mismatched, what tickets are open, and what changed since the last scan.

The server is hosted — no Docker containers, no local proxies, no process to keep running. Connect once via OAuth or API key and every compatible agent gets immediate, scoped access to your project context.

  • 30+ tools covering contracts, schemas, tickets, and scans
  • Hosted endpoint — no local server required
  • OAuth and API key authentication with per-project scoping
  • Compatible with Cursor, Claude, Windsurf, and any MCP client

Pre-analyzed context, not raw diffs

Unlike raw file access or code search, MCP tools return pre-analyzed, contextually relevant information. When an agent asks for mismatches, it receives a prioritized list with severity, affected endpoints, and suggested fixes — not a raw diff it would need to interpret on its own.

The server supports OAuth authentication, session persistence, and per-project access controls. Agents can only see projects their user has access to, and every tool call is logged for auditability.

  • Project summary bootstrap with repo-aware context
  • OpenAPI-aware endpoint inspection and mismatch detection
  • Database schema and type mismatch views
  • Ticket lifecycle, audit trail, and deployment status tools

Built for multi-agent coordination

The same MCP surface works for Cursor, Claude, Lovable, and other compatible clients simultaneously. Multiple agents working on the same project see the same ticket state, the same mismatch list, and the same API contract — eliminating the context drift that typically causes duplicate work or conflicting implementations.

The AppHandoff MCP server supports both Streamable HTTP and SSE transports, so it connects to the full spectrum of MCP-compatible clients without protocol negotiation on your end.

Connect in under five minutes

Add the AppHandoff MCP server endpoint to your client's MCP configuration, authenticate with OAuth or generate an API key from your project settings, and your agent has immediate access to live project context.

The server is always on — backed by two always-running machines — so agents that reconnect mid-session, restart, or run on a schedule pick up exactly where they left off without re-authentication.

  • Single endpoint URL — no port forwarding or local tunnels
  • Session persistence across agent restarts
  • Project-scoped API keys for automated pipelines
  • Streamable HTTP and SSE transport support

FAQ

What is an MCP server?

An MCP server is a service that exposes tools, resources, and prompts to AI models through the Model Context Protocol. It acts as the bridge between an AI coding agent and the external data sources or capabilities the agent needs — in AppHandoff's case, live project contracts, scan results, tickets, and API specifications.

Which MCP clients does AppHandoff support?

Any client that implements the Model Context Protocol. This includes Claude Desktop, Cursor, Windsurf, and the Claude API with tool use. AppHandoff supports both Streamable HTTP and SSE transports, covering the full range of current MCP client implementations.

Do I need to run a local MCP server?

No. AppHandoff's MCP server is fully hosted. There is no Docker container, local process, or tunnel to maintain. Add the endpoint URL to your client configuration, authenticate once, and the server handles the rest.

How is access controlled?

Each user authenticates via OAuth or API key. The server enforces per-project access so agents only see the projects their user is authorised to access. All tool calls are logged for audit purposes.

What tools does the MCP server expose?

The server exposes over 30 tools covering project context (summaries, repo state, scan history), API contract inspection (endpoint specs, mismatch details), database schema views, handoff ticket management (create, update, close tickets), and deployment status. A full reference is available in the documentation.

Related guides

MCP for LovableHow Lovable-built frontends connect to the AppHandoff MCP surface for backend coordination.Lovable to ProductionBridge generated frontends to production repos with contract scanning and handoff tickets.How to set up MCP with LovableStep-by-step guide for connecting AppHandoff's MCP server to a Lovable project.Connect your projectDocumentation for adding the AppHandoff MCP endpoint to Cursor, Claude, or Windsurf.

Next step

See the full workflow in context

The `/v2` buildout keeps the current root app untouched while the future public site takes shape separately.

View pricingCompare approaches