Codex, Cursor, VS Code, Gemini CLI and other agents

Each of these agents reads skills, so one command makes TDC available in all of them. The MCP server is optional on top of it.

The skill

npx -y tdcv2-mcp@0.1.6 install-skill --agents

This puts the skill in ~/.agents/skills/tdcv2 — the shared folder that Codex, Cursor, VS Code with Copilot, and Gemini CLI all read. Start a new session and ask for data in plain words; the agent picks the skill up by itself.

To keep the skill in one repository instead, so everyone who clones it gets it, write it into the repository's own .agents/skills folder and commit it:

npx -y tdcv2-mcp@0.1.6 install-skill --dir .agents/skills

Where each agent looks for skills, according to its own documentation:

Agent	For you	In a repository
Codex	~/.agents/skills	.agents/skills
Cursor	~/.agents/skills, ~/.cursor/skills, ~/.claude/skills	.agents/skills, .cursor/skills, .claude/skills
VS Code (Copilot)	~/.agents/skills, ~/.copilot/skills, ~/.claude/skills	.agents/skills, .github/skills, .claude/skills
Gemini CLI	~/.agents/skills, ~/.gemini/skills	.agents/skills, .gemini/skills
Claude Code	~/.claude/skills	.claude/skills

Running the command again replaces the skill with the release you name, which is how you update it.

The MCP server

The server is the same package, started by the agent rather than by you. Every client takes the same command — npx -y tdcv2-mcp@0.1.6 — and differs only in where the setting goes and what the file looks like.

Codex

Run once:

codex mcp add tdcv2 -- npx -y tdcv2-mcp@0.1.6

Or add the server to ~/.codex/config.toml yourself:

[mcp_servers.tdcv2]
command = "npx"
args = ["-y", "tdcv2-mcp@0.1.6"]

Codex gives a server 10 seconds to start. The first start downloads the package, so run the command in Before the first start once, or raise the limit with startup_timeout_sec = 30 in the same table.

Cursor

Add the server to ~/.cursor/mcp.json to have it in every project, or to .cursor/mcp.json in one project:

{
  "mcpServers": {
    "tdcv2": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "tdcv2-mcp@0.1.6"]
    }
  }
}

VS Code

GitHub Copilot in VS Code reads .vscode/mcp.json in the workspace. The top-level key here is servers, not mcpServers as in the other clients:

{
  "servers": {
    "tdcv2": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "tdcv2-mcp@0.1.6"]
    }
  }
}

To add it to your user profile instead, for every workspace:

code --add-mcp '{"name":"tdcv2","type":"stdio","command":"npx","args":["-y","tdcv2-mcp@0.1.6"]}'

Gemini CLI

Add the server to ~/.gemini/settings.json to have it in every project, or to .gemini/settings.json in one project:

{
  "mcpServers": {
    "tdcv2": {
      "command": "npx",
      "args": ["-y", "tdcv2-mcp@0.1.6"]
    }
  }
}

The MCP server in other clients

Most MCP clients take the same mcpServers entry; look up where yours keeps it:

{ "mcpServers": { "tdcv2": { "command": "npx", "args": ["-y", "tdcv2-mcp@0.1.6"] } } }

On Windows, write "command": "npx.cmd" if the client cannot start npx.

Where the server works

The server treats the folder it is started in as the project: it reads .tdc files and writes generated files only there. Data packs it installs go where the project's tdcv2.config.json says, which is tdcv2-packs/ in the project unless you chose otherwise. If the generated files land somewhere other than your project, tell the client which folder to start the server in: Codex, VS Code and Gemini CLI take a cwd field next to command.

Before the first start

The first start downloads the server and the engine from npm, which can take longer than a client waits. Run the command once in a terminal to have it ready:

npx -y tdcv2-mcp@0.1.6 --help