[Bug] mcp-pandoc 0.8.1 crashes on startup with mcp SDK 2.0.0 (`AttributeError: 'Server' object has no attribute 'list_tools'`)

[pbarone]

What were you trying to accomplish?

Using mcp-pandoc from an MCP client (Kiro IDE) to convert documents between formats (Markdown, DOCX, etc.). The server is launched with uvx mcp-pandoc.

Expected Behavior

The mcp-pandoc server starts, completes the MCP handshake, and exposes the convert-contents tool so the client can connect and run conversions.

Actual Behavior

The server crashes on startup during import, before any MCP handshake, so the client reports it as failing to connect. The crash is an AttributeError raised while defining the tool handlers:

Traceback (most recent call last):
File "", line 198, in _run_module_as_main
File "", line 88, in _run_code
File ".../Scripts/mcp-pandoc.exe/main.py", line 5, in
from mcp_pandoc import main
File ".../site-packages/mcp_pandoc/init.py", line 1, in
from . import server
File ".../site-packages/mcp_pandoc/server.py", line 20, in
@server.list_tools()
^^^^^^^^^^^^^^^^^
AttributeError: 'Server' object has no attribute 'list_tools'

Root cause: mcp-pandoc 0.8.1 does not cap its mcp dependency, so a fresh install resolves the latest mcp Python SDK (2.0.0). SDK 2.0 removed the low-level Server.list_tools() / call_tool() decorator API that server.py relies on, so import fails.

Environment Information

Operating System:
Windows 11 (x64), build 10.0.26100

Python Version:
Python 3.13.1 (uv-provisioned for the mcp-pandoc environment; system Python 3.13.2)

mcp-pandoc Version:
0.8.1 (latest on PyPI)

mcp (Python SDK) Version: 2.0.0 (resolved automatically because it is uncapped)

MCP Client:
Kiro IDE 1.0.242 (also reproduces by running uvx mcp-pandoc directly)

Steps to Reproduce

1. Configure mcp-pandoc with the standard unpinned command: uvx mcp-pandoc.
2. Start the MCP client (or just run uvx mcp-pandoc in a terminal).
3. uvx resolves mcp-pandoc 0.8.1 + mcp 2.0.0, and the process exits immediately with the AttributeError above. The client shows the server as not connecting.

Screenshots or Logs

Verified workaround (pin the SDK back to 1.x):

uvx --from mcp-pandoc==0.8.1 --with "mcp==1.29.0" mcp-pandoc

With mcp 1.29.0 the server initializes and returns its tool:
INIT OK -> serverInfo mcp-pandoc 0.8.1
tools/list -> convert-contents

Suggested fix for the project:

• Short term: cap the dependency to mcp<2 (or mcp>=x,<2) in pyproject.toml and release a patch.
• Longer term: migrate server.py to the mcp SDK 2.0 API and require mcp>=2.

Pre-submission Checklist

• I have searched existing issues to avoid duplicates
• I have provided environment information requested above

[Andrea-encrypted]

I am experiencing the exact same issue on my Windows machine. Looking forward to a fix, thanks for creating this MCP server!

[vivekVells]

@pbarone Thank you for raising this issue and providing the helpful reproduction details. The fix has now been merged and released in mcp-pandoc with version: 0.8.2.

Please update to the latest release using uvx mcp-pandoc@latest or restart your MCP client. Let me know if this resolved your issue @pbarone @Andrea-encrypted.

Follow-up PR change as a long term fix will migrate to MCP SDK v2 while preserving tool schemas, validation, error handling, and client compatibility.

I will keep this bug open for now to follow up.

[vivekVells]

@pbarone @Andrea-encrypted, the immediate fix is available in 0.8.2 PR #41, and PR #42 provides the long-term MCP SDK v2 migration planned for 0.9.0.

Please retry with the latest release. I’ll leave this open for seven days for verification and then close it as resolved if no further problems are reported. If the issue persists, please reply with the updated logs.

[Andrea-encrypted]

Hi @vivekVells, I just updated and tested version 0.8.2 on my Windows machine. The server initialized perfectly without any errors and successfully processed my local Word documents. The fix completely resolved the issue for me. Thank you so much for the incredibly fast response and update!

[vivekVells]

Glad to see you unblocked @Andrea-encrypted. Thanks for confirming. I’m closing #40 as resolved.

@Andrea-encrypted If you’re open to sharing, what broader workflow are you using mcp-pandoc for? Is it for research, document authoring, knowledge management, or something else? Understanding where conversion fits into real workflows helps guide future improvements. Open to your thoughts on where it can help further.