Summary
tigermcp 0.1.7 declares mcp[cli]>=1.13.0 without an upper bound. Now that mcp 2.0.0 is available, a fresh dependency resolution installs MCP 2.0 and TigerMCP fails during startup.
Reproduction
uv run --isolated --with "tigermcp==0.1.7" --with "mcp==2.0.0" \
python -c "from tigermcp.server import main"
Actual behavior
ModuleNotFoundError: No module named "mcp.server.fastmcp"
tigermcp/server.py currently imports:
from mcp.server.fastmcp import FastMCP
That module is no longer present in mcp 2.0.0, so MCP clients only observe the TigerMCP subprocess closing during initialization.
Expected behavior
Installing and running the published tigermcp package should resolve a compatible MCP SDK version and start successfully.
Suggested fix
Until TigerMCP is migrated to the MCP 2.x server API, constrain the dependency in tigeropen/examples/ai/mcp_server/setup.py:
A packaging/startup smoke test using a fresh dependency resolution would also prevent this regression.
Verified workaround
TigerMCP starts successfully when MCP is constrained to 1.x:
uvx --python 3.13 --with "mcp>=1.13,<2" tigermcp
Summary
tigermcp 0.1.7declaresmcp[cli]>=1.13.0without an upper bound. Now thatmcp 2.0.0is available, a fresh dependency resolution installs MCP 2.0 and TigerMCP fails during startup.Reproduction
Actual behavior
tigermcp/server.pycurrently imports:That module is no longer present in
mcp 2.0.0, so MCP clients only observe the TigerMCP subprocess closing during initialization.Expected behavior
Installing and running the published
tigermcppackage should resolve a compatible MCP SDK version and start successfully.Suggested fix
Until TigerMCP is migrated to the MCP 2.x server API, constrain the dependency in
tigeropen/examples/ai/mcp_server/setup.py:"mcp[cli]>=1.13.0,<2",A packaging/startup smoke test using a fresh dependency resolution would also prevent this regression.
Verified workaround
TigerMCP starts successfully when MCP is constrained to 1.x:
uvx --python 3.13 --with "mcp>=1.13,<2" tigermcp