consult-llm CLI
A CLI plus consult, debate and collab skills that give an agent a second opinion from another model inside the current workflow.
- 148 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Files and the diff are sent to another model provider.
- In API mode keys are stored in the local configuration.
Install
- Works in
Manual install
brew install raine/consult-llm/consult-llmFirst Quick Start command; a script install through `curl` also exists. Then run `consult-llm install-skills`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Files and the diff are sent to another model provider.
- In API mode keys are stored in the local configuration.
This is third-party code. Review the repository files before installing.
What it does
consult-llm is a Rust CLI that sends a question to another model and returns the answer into the agent session. A model trained differently pushes back more often and spots blind spots, so a same-model self-check gives way to an independent opinion. Skills install into Claude Code, Codex, OpenCode and Pi with `consult-llm install-skills` and run as `/consult`, `/debate` and `/collab`. Backends come in two kinds: API with a key, and local CLIs such as Gemini CLI or Codex CLI that reuse an existing subscription. It supports multi-turn threads, git diff context, copying a prompt to the clipboard for browser models through `--web`, and a terminal monitor of runs.
Who it is for. Developers on Claude Code or Codex who want an independent review of architecture, a diff or a hard bug from another model without copying by hand.
Good fit when
- A fix did not help and a model with different training should take a look.
- An architecture choice is debated between two models in `/debate` mode.
- A second model reviews a diff while the first keeps working.
- A prompt must move to a browser chat without formatting files by hand.
Not a fit when
- Source code cannot go to an external model provider.
- There are no API keys or subscriptions for other models.
- The task is small and a second opinion costs more than the edit.
Example request
We are stuck on this bug. Ask Gemini through consult-llm, pass the diff and show where the model disagrees.Limitations
Each call sends chosen files and the diff to another model provider, so closed code needs a local backend or a trusted provider. The keyless mode needs a login in the matching CLI, such as `gemini login` or `codex login`. Model lists and names change with releases, and `consult-llm models` prints the current set. The run monitor and binaries target macOS and Linux, and the script install fetches from GitHub with `curl`.
How to disable. Delete the installed skills from the agent directories, then the binaries: `brew uninstall consult-llm` or remove the files from the script install. Clear keys from `consult-llm config`.
FAQ
Which models are available?
The README names GPT, Gemini, Claude, DeepSeek, MiniMax, Grok and GLM, plus any `openrouter/*` model. `consult-llm models` prints the exact list for your version.
Are API keys required?
Not always. The `gemini-cli` and `codex-cli` backends reuse an existing subscription through that CLI login.
How do consult, debate and collab differ?
The three skills run different exchange patterns with another model: a single question, a debate of positions and joint work. The Skills section of the README has details.
How do I send a prompt to a browser chat?
Add the `--web` flag: the formatted prompt is copied to the clipboard and can be pasted into ChatGPT or another web chat.
Where do I see run history?
A terminal monitor shows current and past runs in real time. It ships with `consult-llm`.
Discussion
No questions or reviews yet.
Related
CLI toolsA skill that asks ChatGPT and Gemini before Claude shows its plan, weighs the three views and delivers one plan crediting each model's contribution
A Claude Code plugin that asks several AI models the same question and shows their answers side by side with a synthesis of agreement and divergence
A CLI and manager skill: one skills library distributed to 70+ agents, with management done from the agent conversation
A set of skills that hand a task to a separate CLI agent while you review the diff and land the commit yourself
A skill that links two to four open Codex CLI windows: one splits the work while the others execute, cross-check and deliver.
Skrills validates and syncs skills, subagents, MCP servers and hooks across Claude Code, Codex CLI, Copilot CLI and Cursor
README in short
The tool wraps a call to another model into agent skills. Backend settings live in `consult-llm config`, and a run log is kept separately.