Connecting MCP to T-REX Planner

# Connecting via URL

Use this address if your service simply asks you to provide a link to the MCP server:

http://mcp.trexplanner.com/mcp

# Connecting via JSON configuration

Most tools (Cursor, Claude Desktop, Windsurf, VS Code, etc.) accept configuration in JSON format. Add the block below to your mcp.json config file:

{
  "mcpServers": {
    "trexplanner": {
      "url": "http://mcp.trexplanner.com/mcp"
    }
  }
}

# Choosing available tools

By default, AI gets access to all core TrexPlanner tools except batch command execution — system (Batch).

What is Batch

Batch allows AI to execute up to 20 operations in a single request — for example, editing several tasks at once. Without Batch, changes happen sequentially, one command at a time: this is safer — an error affects only one entity, and execution can be stopped. Batch is disabled by default. Enable it if you trust your assistant.

Available sections

• tasks — reading, creating, and editing tasks and checklists; changing completion status.
• projects — reading, creating, and editing projects.
• meta — core entities: reading and editing.
• habits — reading, creating, and editing habits.
• kanban — reading, creating, and editing kanban boards.
• tags — reading, creating, and editing tags.
• system — batch command execution (Batch).

Toolsets configuration examples

Add the toolsets parameter to your URL and list the desired sections separated by commas.

All tools, including Batch:

http://mcp.trexplanner.com/mcp?toolsets=tasks,projects,meta,habits,kanban,tags,system

Tasks, projects, meta, kanban — with Batch, without habits and tags:

http://mcp.trexplanner.com/mcp?toolsets=tasks,projects,meta,kanban,system

Only tasks and projects, without Batch:

http://mcp.trexplanner.com/mcp?toolsets=tasks,projects

# Instructions for popular services

Choose your tool — step-by-step setup below.

# ChatGPT

MCP connection in ChatGPT is available via the Plugins section. If the + button is not showing in plugins, enable Developer mode — details in the knowledge base.

1. Open ChatGPT and go to Settings → Plugins.
2. Click + → Add by URL and paste the address.
3. Complete OAuth authorization and grant the required permissions.

http://mcp.trexplanner.com/mcp

# Claude Desktop

Claude works with TrexPlanner out of the box and supports built-in smart prompts for planning.

1. Open Claude Desktop.
2. Go to Settings → Connectors.
3. Click Add Connector, choose Server by URL type, and paste the address.
4. Log in with your TrexPlanner account.

# Claude Code

To work directly in the terminal, run this command:

claude mcp add --transport http trexplanner http://mcp.trexplanner.com/mcp

Then run claude, type /mcp and complete OAuth authorization in your browser.

# Codex CLI

To work directly in the terminal, run this command:

codex mcp add trexplanner http://mcp.trexplanner.com/mcp

A browser window will automatically open for OAuth authorization and permission confirmation.

# Cursor / Windsurf / VS Code

In modern AI-powered IDEs, TrexPlanner can act as a built-in task tracker: move tasks on a kanban board, create bug reports, and update checklists without leaving your editor.

1. Open your IDE settings.
2. Find the MCP config file: mcp.json in Cursor/Windsurf or .vscode/mcp.json for VS Code.
3. Add the block below:

{
  "mcpServers": {
    "trexplanner": {
      "url": "http://mcp.trexplanner.com/mcp"
    }
  }
}

Restart your IDE and complete OAuth authorization in the browser.

# DeepSeek

The DeepSeek chat interface does not yet support direct MCP connections. Connection is only possible through the DeepSeek API with an additional integration layer on your side.

How to connect via API

1. Create an API key in the DeepSeek platform.
2. Set up data exchange between DeepSeek API and the MCP server through your own proxy service or an automation platform (n8n, Make, custom backend).
3. The proxy must:
   • accept requests from DeepSeek in chat-completions format;
   • call TrexPlanner MCP tools at http://mcp.trexplanner.com/mcp;
   • return results back in DeepSeek format.

Useful links:

• Create DeepSeek API keys → API Keys section in your account.
• DeepSeek API docs → platform.deepseek.com/docs.

# Alice AI & Yandex neural networks

The Alice AI chat interface doesn't support direct MCP connections yet, but you have two working options.

Option 1 — TrexPlanner skill for Alice

Connect the official TrexPlanner skill in the Yandex app. This is the fastest way to manage tasks by voice and chat with Alice. Instructions are in the knowledge base: "Yandex Alice Integration".

Option 2 — Yandex neural networks via MCP

If you want to connect Yandex Cloud / AI Studio neural networks to MCP, technical setup on your side is required. This scenario is for developers.

• Register in Yandex Cloud and activate the AI Studio service.
• Create an agent or scenario and add the MCP server with URL http://mcp.trexplanner.com/mcp to the available tools.
• Complete authorization and grant the required permissions.

Useful links:

• What is MCP in Yandex Cloud → Yandex Cloud documentation.
• MCP Hub in Yandex AI Studio → "Tools / MCP" section in AI Studio.

# GigaChat

GigaChat does not support simple MCP connection through its regular chat interface. Connection is possible through technical setup — locally via LangChain / GigaChain or remotely via HTTP/SSE.

Local scenario (LangChain / GigaChain)

1. Install dependencies: pip install gigachain langchain mcp-client.
2. Create an MCP client and connect to the TrexPlanner server.
3. Pass the MCP tools list to a GigaChain agent and register them as functions.

from gigachat import GigaChat
from mcp import MCPClient

# 1. Connect to the TrexPlanner MCP server
mcp = MCPClient("http://mcp.trexplanner.com/mcp")
tools = mcp.list_tools()

# 2. Initialize GigaChat with your API key
giga = GigaChat(credentials="YOUR_API_KEY")

# 3. Pass MCP tools to a GigaChain agent
# (pseudocode: actual assembly depends on your GigaChain version)
agent = giga.create_agent(tools=tools)

response = agent.run("Create a task for tomorrow: prepare a report")
print(response)

Remote scenario (HTTP / SSE)

1. Deploy an HTTP/SSE wrapper over the TrexPlanner MCP server on your own server.
2. Register the endpoint as a tool in your GigaChat agent.
3. Handle OAuth authorization and forward calls to MCP.

Useful links

• Example MCP connection via GigaChain → "MCP / Tools" section in GigaChain documentation.
• GigaChat documentation → "Agents & Function Calling" section.

# Other services

If you use LM Studio, Jan, Open WebUI, LibreChat, or other MCP-compatible clients — simply add the JSON block from the "Connecting via JSON configuration" section to the appropriate config file, or enter the server URL in the "Add MCP server" form. For fine-grained permission control, use the toolsets parameter from the "Choosing available tools" section.