Documentation
MCP

MCP Overview

Model Context Protocol (MCP) lets Flapjack agents use external tools from GitHub, Slack, Supabase, and any MCP-compatible server.

The Model Context Protocol (MCP) is an open standard for connecting AI agents to external tools and data sources. Flapjack supports MCP natively — connect any MCP-compatible server and your agents can use its tools during conversations.

What MCP Enables

With MCP, your Flapjack agents can:

  • Query GitHub — list repos, create issues, search code
  • Send Slack messages — post to channels, read threads
  • Query databases — via Supabase, Neon, or custom MCP servers
  • Access any API — any service with an MCP server

How It Works

User sends message
    → Agent has MCP tools in its tool list
    → Agent decides to call an MCP tool
    → Flapjack calls the MCP server
    → Server executes the tool and returns results
    → Agent incorporates results into its response

Transport Types

MCP servers can connect via three transport types:

TransportProtocolUse Case
streamable_httpHTTP POST with streamingRemote servers (most common)
sseServer-Sent EventsLegacy remote servers
stdioStandard I/O subprocessLocal command-line tools

Tool Namespacing

When an MCP server is connected, its tools are namespaced with the server's slug to avoid name collisions:

{slug}__{tool_name}

For example, a GitHub MCP server with slug github and a tool named list_repos becomes github__list_repos.

Credentials

MCP server credentials are encrypted at rest using AES-256-GCM. Flapjack supports:

Credential TypeDescription
Environment variablesKey-value pairs passed to the server
HTTP headersHeaders included in requests
OAuth tokensOAuth 2.1 tokens with automatic refresh

MCP Registry

Flapjack integrates with the official MCP registry at registry.modelcontextprotocol.io. You can search for available servers directly from the dashboard or API.

Sections

Next Steps

Docs last updated May 11, 2026