LLM CLI Bridge is an MCP server that lets Claude orchestrate Gemini CLI or Codex CLI for tasks that benefit from larger context windows, backend-specific strengths, or multi-round review workflows.

Why This Exists

Claude is strong at orchestration and reasoning. Gemini is useful when you want very large context handling. Codex is useful when you want GPT-based code review and coding workflows. This tool gives Claude a single MCP surface that can delegate to either backend without forcing you to change your workflow.

Tools

ask

Query Gemini or Codex with @ file references. Supports sessions, change mode, and backend selection.

text
ask gemini to analyze @src/main.js
ask codex to analyze @src/main.js
use codex with high reasoning to review @package.json

brainstorm

Structured ideation with multiple methodologies and optional backend selection.

Auto - Let the tool pick the right approach
Divergent - Generate many ideas quickly
Convergent - Narrow and refine options
SCAMPER - Systematic prompt-based ideation
Design Thinking - Human-centered exploration
Lateral - Unexpected angles and connections

review-code

Interactive, multi-round code review with git-aware sessions and comment decision tracking.

Multiple review types: general, security, performance, quality, architecture
Severity filters for tighter review passes
Follow-up rounds with accept/reject/modify/defer decisions

Installation

bash
# Recommended
claude mcp add llm-cli -- npx -y @maxanatsko/llm-cli-bridge@latest

# Verify installation
/mcp

Claude Desktop Configuration

json
{
  "mcpServers": {
    "llm-cli": {
      "command": "npx",
      "args": ["-y", "@maxanatsko/llm-cli-bridge@latest"]
    }
  }
}

Backends

LLM CLI Bridge supports both backend families:

Gemini CLI - large-context analysis and research workflows
Codex CLI - GPT-based code review and coding workflows

You can switch backends at the prompt level instead of maintaining separate MCP servers.

Repository

View the source code on GitHub or install from npm.