Consult7 is an MCP server that enables AI agents to offload analysis of large file collections to high-capacity language models via OpenRouter, supporting context windows up to 2M tokens. It collects files from specified paths, assembles them into a single context, sends them to a selected model (such as Google's Gemini 3.1 family or other frontier models) with a user-provided query, and returns the results back to the agent. This solves the problem of agents running out of context when analyzing large codebases, document repositories, or mixed content that exceed their current limits, while also enabling access to specialized model capabilities and comparative analysis across different models.
claude mcp add --transport stdio szeider-consult7 uvx consult7