MoonshotAI / MoonshotAI/kimi-cli

MCP startup UI stays at `0/5 connected` even though all servers are connected and usable

Open
#2,328 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
11.4k
Forks
1.3k
Avg merge
9h 47m
Merged PRs (30d)
2

Description

Summary

On Linux Mint 22.3, Kimi CLI startup UI keeps showing:

  • MCP Servers: 0/5 connected, 0 tools
  • connecting to mcp servers...

But the MCP servers are actually connected successfully, handshake completes, and tool calls work.

Environment

  • Kimi CLI: 1.43.0
  • OS: Linux Mint 22.3
  • Working directory: /data/codeRepository/desmart/git/mdmNew
  • Custom local MCP servers over stdio

What happens

At startup, the terminal UI remains in a connecting state and still displays 0/5 connected, but backend logs show all MCP servers connected normally.

Evidence

1. Kimi CLI backend log shows all 5 MCP servers connected

From ~/.kimi/logs/kimi.log for session 24fa57d2-37ef-4ea6-8371-307a5a1d0ee9:

  • Connected MCP server: mdm-file-parser
  • Connected MCP server: mdm-git
  • Connected MCP server: mdm-mysql-metadata
  • Connected MCP server: mdm-fetch
  • Connected MCP server: mdm-code-intelligence
2. MCP protocol trace shows successful handshake

For all 5 servers, the following sequence completes successfully:

  • initialize
  • notifications/initialized
  • tools/list
3. Real tool calls succeed

Using the same ~/.kimi/mcp.json config that Kimi CLI reads:

  • mdm-git.branch_list returns branch data successfully
  • mdm-code-intelligence.get_mcp_health returns loading first and then ready

Example final health result from mdm-code-intelligence:

{"status":"ready","index_ready":true,"build_in_progress":false,"total_nodes":35725}

Expected behavior

After MCP servers are connected, the startup UI should reflect the actual status instead of staying at 0/5 connected / connecting to mcp servers....

Actual behavior

The UI appears stuck in the connecting state, even though the logs and real tool calls show MCP is already working.

Notes

I have already ruled out local MCP server issues on my side:

  • all servers complete stdio MCP handshake
  • tools/list succeeds
  • tool calls succeed
  • startup stderr noise has been removed
  • server launch config is direct script execution (not a bootstrap wrapper)

So this looks more like a Kimi CLI UI/state refresh issue than an MCP server connection failure.

If helpful, I can also provide a minimal reproduction trace of the MCP request/response sequence.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the startup using the ~/.kimi/mcp.json configuration and compare the UI with ~/.kimi/logs/kimi.log, where the five successful connections are recorded. Trace how MCP handshake and tools/list results update the startup connection count and status, then verify that the UI reaches the connected state while real tool calls still succeed.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
cli
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.