Quick Start › Integrations › Other MCP clients (generic)

Annex — Other MCP clients (generic)

Generic guidance for any MCP-capable client without a dedicated annex

This guide is for any AI client with MCP support that does not have a dedicated guide in this manual: Kimi Code, Qwen Code, Cursor, Windsurf, VS Code MCP extensions, other existing IDEs, and the ones that keep appearing.

Use a dedicated guide first when one exists. Those pages contain client-specific behavior, tested workarounds, and exact UI names. This generic guide covers the common MCP mechanics only.

Recent examples include Kimi Code with Kimi K3 and Qwen Code with Qwen3.7. Multi-model clients may also expose models such as MiniMax M3, LongCat-2.0, DeepSeek V4, and MiMo V2.5. The model name does not change KaKeKiKoKu's two requirements: a local STDIO MCP connection and persistent startup instructions in the file that client loads.

Honesty up front. KaKeKiKoKu does not test every MCP client on the market. This page gives you the safe pattern for unvalidated clients, plus the troubleshooting moves learned from the validated guides.

KaKeKiKoKu works best in these clients in Precise mode: the AI can call KaKeKiKoKu MCP tools, while KaKeKiKoKu UI actions that would push messages to the AI copy text to the clipboard for you to paste.

Requirements

Before you start the KaKeKiKoKu integration flow, you need:

The MCP transport spec defines STDIO as a client launching a server subprocess and exchanging JSON-RPC over stdin/stdout. That is the pattern KaKeKiKoKu uses.

macOS note. KaKeKiKoKu does not ship a public macOS build today. macOS clients are future scope until KaKeKiKoKu itself ships for macOS.

↑ Back to top

1. Register KaKeKiKoKu in your client

The KaKeKiKoKu command is OS-specific. Use the one that matches the machine running the AI client.

wuniq-mcp.exe

Linux. Install KaKeKiKoKu from the Snap Store first:

sudo snap install wuniq
/snap/bin/wuniq.wuniq-mcp

Almost every MCP client fits one of these patterns. Pick the one your client exposes.

Pattern A — Configuration UI panel

The client has an MCP servers screen with a form. Common fields are:

Windows path note. Use the alias wuniq-mcp.exe. Do not point the client at an absolute path under C:\Program Files\WindowsApps\...; Windows protects that folder and external MCP clients can hit Permission denied or EPERM.

Pattern B — JSON or TOML configuration file

The client reads a local configuration file. If the file already contains other MCP servers, keep them and add wuniq beside them.

JSON shape

{
  "mcpServers": {
    "wuniq": {
      "command": "wuniq-mcp.exe"
    }
  }
}
{
  "mcpServers": {
    "wuniq": {
      "command": "/snap/bin/wuniq.wuniq-mcp"
    }
  }
}

TOML shape

[mcp_servers.wuniq]
command = "wuniq-mcp.exe"
enabled = true
[mcp_servers.wuniq]
command = "/snap/bin/wuniq.wuniq-mcp"
enabled = true

Pattern C — CLI command

Some clients offer a command to register MCPs without touching files. The exact syntax varies, but it usually looks like one of these:

<client> mcp add wuniq -- wuniq-mcp.exe
<client> mcp add wuniq -- /snap/bin/wuniq.wuniq-mcp

To verify registration, many clients expose a list command:

<client> mcp list

Check your client's --help for the exact syntax.

↑ Back to top

2. Startup instructions (the second piece of the integration)

This is the second of the two pieces every integration needs. MCP registration gives the AI the KaKeKiKoKu tool; startup instructions tell it to use KaKeKiKoKu at the beginning of a session.

Many clients read one of these files from the project root:

If your client respects any of them, put this block in the file it loads:

# KaKeKiKoKu Studio (Knowledge Engine)

At the start of every session, run `open list` using the KaKeKiKoKu MCP tool. If it returns projects, `open` each one. If the list is empty, proceed normally — the user may open projects later from the KaKeKiKoKu UI. Never assume the working directory is a KaKeKiKoKu project; the working directory may be a container folder holding multiple projects. Only open paths that `open list` explicitly returns.

Before working on any code file, `read` its KaKeKiKoKu sidecar for context. After changing code, run `sync`.

If your client uses another rule file. Keep the same KaKeKiKoKu block and place it wherever that client loads project instructions.

If your client has no project-instruction mechanism, you will need to ask the model manually at the start of each conversation:

Run open list with the KaKeKiKoKu tool, then open every project it returns.

↑ Back to top

3. Restart the client

After adding or editing MCP configuration and startup instructions, close the AI client and reopen it. Then start a fresh conversation. Most clients load MCP servers and rule files at session startup.

↑ Back to top

4. Open a project and verify

  1. Open the client on your KaKeKiKoKu working directory or on a project folder inside it.
  2. Start a fresh conversation.
  3. If your startup instructions are loaded, the AI should call KaKeKiKoKu automatically. If not, ask:

    Run open list with the KaKeKiKoKu tool.

  4. If everything is connected, the AI will list the projects KaKeKiKoKu knows about.
  5. Ask it to open the project you want, or let the startup instruction open every returned project.
  6. In the KaKeKiKoKu UI, the AI-connected indicator should turn green when the client uses the MCP tool.

Quick test

Read the project index with KaKeKiKoKu.

The AI should use the wuniq MCP tool. If it answers from general memory, ask explicitly to use KaKeKiKoKu and re-check the startup instructions file.

↑ Back to top

5. Working in Precise mode

Assume Precise mode for generic MCP clients unless a dedicated KaKeKiKoKu guide says otherwise. In Precise mode:

Recommended KaKeKiKoKu configuration. Go to KaKeKiKoKu > Settings > Application > AI Sync and choose Precise mode.

↑ Back to top

6. Ask the client's own AI

If you cannot find where your client configures MCP, ask the client's own AI. It often knows its own UI, config file, and current version better than generic third-party documentation.

Try these prompts:

How do I register a local STDIO MCP server named wuniq in this IDE?

Where is this client's MCP configuration file, and what JSON or TOML shape does it expect?

Which project instruction file does this IDE read at the start of a conversation?

Paste the KaKeKiKoKu command for your OS when you ask. If the AI gives a config shape, compare it with the patterns above and keep your existing MCP entries.

↑ Back to top

7. Universal troubleshooting

KaKeKiKoKu doesn't appear as an available tool

  1. Verify the client registered the MCP correctly (check the panel, the JSON/TOML, or <client> mcp list as appropriate).
  2. Verify the command matches your OS.
  3. Close and reopen the client. Most load MCP config only at startup.
  4. If it still doesn't appear, ask the client's AI how to verify which MCPs are connected.

Config file saved with a BOM

If you used Pattern B (a JSON or TOML config file) and the client never loads KaKeKiKoKu, check the file's encoding. Some editors — classic Notepad's UTF-8 option, or PowerShell's Set-Content -Encoding UTF8 — prepend a hidden 3-byte marker (a UTF-8 BOM) at the start of the file. Strict parsers reject it; a Go-based parser, for example, fails with invalid character 'ï' looking for beginning of value. The marker is invisible in the editor and online validators usually do not catch it. Re-save the file as UTF-8 without BOM — most editors show and let you change the encoding in the status bar.

"Permission denied", "Access denied" or "EPERM" on Windows

Typical cause: you configured an absolute path inside C:\Program Files\WindowsApps\.... That folder is protected by Windows. Fix: change the command to the alias wuniq-mcp.exe.

where.exe wuniq-mcp.exe returns nothing

Windows can't find the alias. Check:

  1. That KaKeKiKoKu is installed.
  2. That you can open KaKeKiKoKu from the Start menu.
  3. That your PATH includes %LOCALAPPDATA%\Microsoft\WindowsApps.

If you just installed KaKeKiKoKu, close and reopen your terminal.

Linux command is not found

Check that the KaKeKiKoKu Snap is installed and that this command exists:

/snap/bin/wuniq.wuniq-mcp

Confirm you installed KaKeKiKoKu in the same Linux user account that runs the AI client.

KaKeKiKoKu can't see projects outside /home/ (Linux)

The KaKeKiKoKu snap is sandboxed by default and cannot read files outside your home folder unless you grant that permission. If your projects live in /mnt/, /media/, a USB drive, or a VM shared folder such as /mnt/hgfs/, run:

sudo snap connect wuniq:removable-media

Then reload the KaKeKiKoKu web UI and open the project again.

Transport closes or the MCP crashes during tool calls

Some clients mishandle MCP server diagnostic output. The MCP spec allows servers to write logs to stderr, but not every client handles that equally well. If you see repeated transport failures, try a quiet wrapper.

Create:

C:\Users\<you>\wuniq-mcp-quiet.cmd

With:

@echo off
wuniq-mcp.exe %* 2>nul

Then configure the client to run the wrapper instead of wuniq-mcp.exe.

Create:

~/wuniq-mcp-quiet.sh

With:

#!/usr/bin/env bash
exec /snap/bin/wuniq.wuniq-mcp "$@" 2>/dev/null

Make it executable:

chmod +x ~/wuniq-mcp-quiet.sh

Then configure the client to run the wrapper instead of the direct KaKeKiKoKu command.

The client starts the MCP but the AI doesn't use the tool

  1. Make sure external tools / MCP tools are enabled in the conversation (some clients have toggles).
  2. Ask explicitly: "Use the wuniq tool to read context before answering."
  3. Consider using startup instructions so the AI invokes KaKeKiKoKu without you having to remind it every time.

Still stuck?

Email support@kakekikoku.com with:

KaKeKiKoKu can also generate a diagnostic report (Support function in the app) that captures the infrastructure state — attach it to the email.

↑ Back to top

8. If you connected a client that's not covered here

If you managed to register KaKeKiKoKu in an MCP client without a dedicated guide and everything works reasonably well, get in touch with:

With enough feedback, KaKeKiKoKu may add a dedicated guide in a future version of the manual.

↑ Back to top