Talk to AI.
Create together.

CODEX

Using Codex

Set this up in the Windows desktop app.

1. Open MCP settings

   Open “Settings” → “MCP servers” → “Add server.”

2. Enter these values

   Name

   cocochara

   Transport

   STDIO

   Command

   Paste the original path copied from COCO CHARA. Do not add quotation marks around it.

   Arguments and environment variables

   Leave these empty.

   STDIO starts the connection program on your PC.

3. Save and restart

   Save the server, select “Restart,” and check that cocochara is enabled in the server list.

4. Check the connection

   Open a local task running on this PC. Type /mcp in the composer to inspect connections, then follow the connection check below.

Settings reference: OpenAI’s MCP guide

CLAUDE DESKTOP

Using Claude Desktop

Use Chat in the installed desktop app.

1. Open the configuration file

   From the app menu, open “Settings” → “Developer” → “Edit Config.” If a folder opens, right-click claude_desktop_config.json and choose “Open with” → “Notepad.”

   You can also enter %APPDATA%\Claude in File Explorer’s address bar. “Edit Config” creates the file if needed.

2. Paste and save

   If the file is empty or contains only {}, select and copy the entire example below, then paste it into the file open in Notepad. Otherwise, expand the instructions for an existing file.

   Example to paste into claude_desktop_config.json

   {
     "mcpServers": {
       "cocochara": {
         "command": "REPLACE_WITH_YOUR_MCP_SERVER_PATH"
       }
     }
   }

   If your file already contains settings

   Keep your existing settings. Add the entry below inside the existing mcpServers braces, after the last server. Put a comma after that server’s closing }, then paste this entry. If mcpServers is empty, paste the entry inside its braces without a preceding comma. If cocochara already exists, update its command only. Do not add a second mcpServers block.

   Entry to add inside mcpServers in claude_desktop_config.json

     "cocochara": {
       "command": "REPLACE_WITH_YOUR_MCP_SERVER_PATH"
     }

   If there are other settings but no mcpServers entry, add "mcpServers": { … } inside the outermost braces, separated from the preceding setting by a comma. Save a copy of the original file before editing.

   Next, replace the path in the file you are editing. Return to COCO CHARA’s AI Settings and copy “MCP Server Path” again. Back in claude_desktop_config.json, select only REPLACE_WITH_YOUR_MCP_SERVER_PATH and paste your path. Keep the surrounding double quotes.

   Inside this JSON file, write every backslash \ twice as \\. For example, a copied path of C:\Games\COCO CHARA\mcp_server\cocochara_mcp.exe becomes "command": "C:\\Games\\COCO CHARA\\mcp_server\\cocochara_mcp.exe". This location is only an example; use your actual copied path.

   Press Ctrl + S. Keep the name claude_desktop_config.json, with no .txt suffix.

3. Quit and reopen Claude Desktop

   Right-click Claude in the Windows notification area (under “∧” if hidden) and quit. Open Claude again from the Start menu.

4. Find the connection

   In a new chat, open “+” → “Connectors” and look for cocochara. If missing, check “Developer” for connection errors. Continue with the connection check.

This local configuration works in desktop Chat; it is not available in the browser or Cowork.

Official guides: Local setup · Connection status · Desktop, web, and Cowork

ANTIGRAVITY

Using Antigravity

These steps use Antigravity IDE, with its editor and AI side panel.

1. Open MCP management

   At the top of the AI side panel, select “…” → “MCP Servers” → “Manage MCP Servers” → “View raw config.”

2. Add the settings

   In mcp_config.json, select and copy the entire example below, then paste it into the empty file or replace {} in the Antigravity editor. For an existing configuration, follow the expandable instructions to add cocochara.

   Example to paste into mcp_config.json

   {
     "mcpServers": {
       "cocochara": {
         "command": "REPLACE_WITH_YOUR_MCP_SERVER_PATH"
       }
     }
   }

   If your file already contains settings

   Keep your existing settings. Add the entry below inside the existing mcpServers braces, after the last server. Put a comma after that server’s closing }, then paste this entry. If mcpServers is empty, paste the entry inside its braces without a preceding comma. If cocochara already exists, update its command only. Do not add a second mcpServers block.

   Entry to add inside mcpServers in mcp_config.json

     "cocochara": {
       "command": "REPLACE_WITH_YOUR_MCP_SERVER_PATH"
     }

   If there are other settings but no mcpServers entry, add "mcpServers": { … } inside the outermost braces, separated from the preceding setting by a comma. Save a copy of the original file before editing.

   Next, replace the path in the file you are editing. Return to COCO CHARA’s AI Settings and copy “MCP Server Path” again. Back in mcp_config.json, select only REPLACE_WITH_YOUR_MCP_SERVER_PATH and paste your path. Keep the surrounding double quotes.

   Inside this JSON file, write every backslash \ twice as \\. For example, a copied path of C:\Games\COCO CHARA\mcp_server\cocochara_mcp.exe becomes "command": "C:\\Games\\COCO CHARA\\mcp_server\\cocochara_mcp.exe". This location is only an example; use your actual copied path.

   Press Ctrl + S.

3. Reopen the app

   Quit and restart Antigravity. Check that cocochara appears and is enabled in MCP management.

4. Try it in chat

   Start a new conversation and follow the connection check.

Interface reference: Google’s Antigravity MCP guide

LM STUDIO

Using LM Studio

Use LM Studio’s own chat with a model loaded on your PC.

1. Prepare a model

   If needed, download a model from “Discover.” Choose one that fits your PC’s memory and supports tool use.

   Open “Chat,” select the downloaded model in the model loader at the top, and load it. Send a short greeting to check that it responds.

2. Open MCP settings

   In the chat’s right settings sidebar, choose “Program” → “Install” → “Edit mcp.json.”

3. Paste and save

   Select and copy the entire example below, then paste it into the empty mcp.json or replace {} in LM Studio’s editor. For an existing file, follow the expandable instructions to add cocochara.

   Example to paste into mcp.json

   {
     "mcpServers": {
       "cocochara": {
         "command": "REPLACE_WITH_YOUR_MCP_SERVER_PATH"
       }
     }
   }

   If your file already contains settings

   Keep your existing settings. Add the entry below inside the existing mcpServers braces, after the last server. Put a comma after that server’s closing }, then paste this entry. If mcpServers is empty, paste the entry inside its braces without a preceding comma. If cocochara already exists, update its command only. Do not add a second mcpServers block.

   Entry to add inside mcpServers in mcp.json

     "cocochara": {
       "command": "REPLACE_WITH_YOUR_MCP_SERVER_PATH"
     }

   If there are other settings but no mcpServers entry, add "mcpServers": { … } inside the outermost braces, separated from the preceding setting by a comma. Save a copy of the original file before editing.

   Next, replace the path in the file you are editing. Return to COCO CHARA’s AI Settings and copy “MCP Server Path” again. Back in mcp.json, select only REPLACE_WITH_YOUR_MCP_SERVER_PATH and paste your path. Keep the surrounding double quotes.

   Inside this JSON file, write every backslash \ twice as \\. For example, a copied path of C:\Games\COCO CHARA\mcp_server\cocochara_mcp.exe becomes "command": "C:\\Games\\COCO CHARA\\mcp_server\\cocochara_mcp.exe". This location is only an example; use your actual copied path.

   Press Ctrl + S. Saving loads the configured servers.

4. Try the connection in chat

   Check cocochara in “Program” and enable it if disabled. Send the connection check below in the chat with your loaded model.

   If asked to approve a tool call, check that it inspects the connection and allow it. If the model only replies without checking, review tool availability and the model’s tool-use support.

Official guides: MCP setup · Loading a model · Model tool support · Loading servers and approving calls