Ask Llm
One MCP entry that auto-detects Gemini, Codex, and Ollama CLIs and exposes whichever tools are installed.
Overview
io.github.Lykhoyda/ask-llm is a MCP server for the Build phase that auto-detects Gemini, Codex, and Ollama CLIs and registers matching tools for your agent over stdio.
What is this MCP server?
- Unified ask-llm-mcp server (version 0.3.8) with stdio npm transport
- Auto-detects Gemini CLI, Codex CLI, and Ollama—registers only available backends
- OLLAMA_HOST default http://localhost:11434 for local models
- Shared GMCPT_TIMEOUT_MS (300000 ms) and GMCPT_LOG_LEVEL across CLIs
- Replaces juggling separate ask-gemini, ask-codex, and Ollama configs when multiple are installed
- Server version 0.3.8
- Detects up to 3 backend families: Gemini, Codex, Ollama
- Default OLLAMA_HOST http://localhost:11434
Community signal: 10 GitHub stars.
What problem does it solve?
Maintaining separate MCP configs per LLM CLI breaks quickly when you only run two of three tools or swap between local Ollama and cloud CLIs.
Who is it for?
Builders experimenting with multiple LLM CLIs plus optional local Ollama who want one MCP integration point.
Skip if: Setups that need only one fixed backend (use the dedicated ask-codex or ask-gemini server instead) or pure cloud API MCP with no local CLI.
What do I get? / Deliverables
After installing ask-llm-mcp, your agent sees a single server whose tool list reflects whichever Gemini, Codex, and Ollama interfaces are actually present on the machine.
- Dynamic MCP tool list reflecting installed LLM CLIs
- Ollama connectivity via configurable OLLAMA_HOST
- Shared timeout and logging behavior across detected backends
Recommended MCP Servers
Journey fit
How it compares
Unified MCP detector for several CLIs, not separate per-vendor servers and not an agent skill playbook.
Common Questions / FAQ
Who is io.github.Lykhoyda/ask-llm for?
Solo developers wiring Claude Code or Cursor to every LLM CLI they keep installed, including optional Ollama on localhost.
When should I use io.github.Lykhoyda/ask-llm?
Use it during Build when you integrate agent tooling, or anytime you want one MCP server that adapts to Gemini, Codex, and Ollama availability.
How do I add io.github.Lykhoyda/ask-llm to my agent?
Run ask-llm-mcp via npm stdio in your MCP config, install the CLIs you need, set OLLAMA_HOST if Ollama is not on default port 11434, and tune GMCPT_TIMEOUT_MS and GMCPT_LOG_LEVEL.