MCP Install Matrix

This matrix documents practical local setup paths for Claude, ChatGPT, Gemini, and Copilot.

The Neksus MCP server itself is local stdio:

neksus-jobspec-mcp

Install dependency extra first:

pip install "neksus-jobspec[mcp]"

Claude

Direct local MCP is supported in Claude Desktop.

Use your Claude MCP/local server config to run:

• command: neksus-jobspec-mcp
• args: []

Reference: Anthropic local MCP setup docs. Anthropic: local MCP on Claude Desktop

ChatGPT

ChatGPT supports MCP through custom connectors/developer mode flows, but local process wiring can vary by plan/workspace and product surface.

Recommended path:

1. Use MCP connector workflow where available.
2. If direct local process wiring is unavailable in your environment, use:
3. the Neksus CLI locally, and
4. the ChatGPT prompt pack in skills/chatgpt/.

Reference: OpenAI MCP and connectors documentation. OpenAI: MCP docs and OpenAI Help: connectors in ChatGPT

Gemini

Gemini CLI supports MCP server configuration, including stdio command-based servers.

Configure your Gemini MCP settings to launch:

• command: neksus-jobspec-mcp
• args: []

If your Gemini environment does not expose MCP config controls, use CLI + prompt-pack fallback (skills/gemini/).

Reference: Gemini CLI MCP documentation. Gemini CLI MCP docs

Copilot

Copilot MCP capability depends on editor/client and policy controls (especially in Business/Enterprise contexts).

When MCP server configuration is available, point it to:

• command: neksus-jobspec-mcp
• args: []

If local MCP server configuration is not available in your Copilot environment, use CLI + prompt-pack fallback (skills/copilot/).

Reference: GitHub MCP docs for Copilot. GitHub Docs: About MCP in Copilot

Fallback path (all assistants)

When direct MCP wiring is unavailable:

1. Run neksus-jobspec commands locally.
2. Use assistant prompt packs under skills/.
3. Keep workflows local-first with validated YAML and deterministic exports.