Graphify-Labs/graphify · error · ImportError

mcp not installed. Run: pip install "graphifyy[mcp]"

Error message

mcp not installed. Run: pip install "graphifyy[mcp]"

What it means

Raised in _build_server (graphify/serve.py) when `from mcp.server import Server` (or `from mcp import types`) fails with ImportError. Every transport (stdio and HTTP) builds its tool/resource registrations through this function, so without the mcp package no server can start. The chained original ImportError identifies exactly which module was missing.

Source

Thrown at graphify/serve.py:1459

        clean = sanitize_label(str(community_name))
        if clean and clean != base:
            return f"{base} — {clean}"
    return base


def _build_server(graph_path: str):
    """Build the configured low-level MCP Server (shared by every transport).

    All graph query tools and resources are registered here over a single
    ``mcp.server.Server`` instance; the caller picks the transport (stdio or
    Streamable HTTP) and runs it. Hot-reload of graph.json works the same way
    regardless of transport, since reloads happen inside the tool handlers.
    """
    try:
        from mcp.server import Server
        from mcp import types
    except ImportError as e:
        raise ImportError('mcp not installed. Run: pip install "graphifyy[mcp]"') from e
    try:
        from mcp.types import AnyUrl
    except ImportError:
        # mcp >= 2.0 dropped the AnyUrl re-export; it was always pydantic's
        # AnyUrl (pydantic is an mcp dependency, so this import cannot miss).
        from pydantic import AnyUrl

    from graphify import paths as _paths

    # Graph contexts comprise one pinned configured default plus a bounded LRU
    # of project_path graphs. This preserves the configured graph's warm index
    # while preventing a shared server from retaining every project it serves.
    _default_graph_path = str(Path(graph_path).resolve())
    _ctx_cache = _GraphContextCache(_max_server_contexts())

    def _load_ctx(path: str):
        """Return the current default or project graph context as a tool error.

View on GitHub (pinned to 7fe58b0b0f)

Solutions

1. Install the extra: pip install "graphifyy[mcp]" (message is literal).
2. Verify the interpreter matches: `python -m pip install "graphifyy[mcp]"` using the same python that runs the server.
3. Confirm the import works: python -c "from mcp.server import Server; from mcp import types".
4. If the import still fails after install, check for a conflicting local file/dir named mcp/ shadowing the package.

Example fix

# before
pip install graphifyy
graphify serve  # ImportError: mcp not installed

# after
pip install "graphifyy[mcp]"
graphify serve

def mcp_extra_installed() -> bool:
    try:
        import mcp.server, mcp.types  # noqa: F401
    except ImportError:
        return False
    return True

assert mcp_extra_installed(), 'pip install "graphifyy[mcp]"'

def can_build_mcp_server() -> bool:
    return mcp_extra_installed()

try:
    _build_server(graph_path)
except ImportError as e:
    if 'graphifyy[mcp]' in str(e):
        sys.exit('Missing MCP dependencies. Run: pip install "graphifyy[mcp]"')
    raise

When it happens

Trigger: Calling serve()/serve_http or otherwise triggering _build_server in an environment where the `mcp` package (the pip extra graphifyy[mcp]) is not installed. Also fires when mcp is installed but broken — wrong version, missing dependency, partially uninstalled.

Common situations: Installing graphifyy without extras (pip install graphifyy) and then running graphify serve; a venv mismatch where the server runs under a different interpreter than the one with mcp; dependency resolver downgrading/removing mcp.

Related errors

• HTTP transport needs the mcp extra (mcp + starlette + uvicor
• graphify install is incomplete: missing always-on block '{ba
• the '{pkg}' package is required for this backend but is not
• the 'anthropic' package is required for this backend but is
• Claude Code CLI not found on $PATH. Install from https://cla

AI-assisted analysis of Graphify-Labs/graphify@7fe58b0b0f (2026-08-14). Data as JSON: /api/errors/9e2be854a178d32c. Report an issue: GitHub.