Skip to main content
Research source in Claude Octopus Claude Octopus runs research as several model perspectives in parallel. With OCTOPUS_RESEARCH_WEB_SEARCH=keenable, it first runs one Keenable search for the research question and gives every perspective the top five results as leads. It works without an API key. → Optional web search suggestions in Claude Octopus’s research docs

Install

The option ships in Claude Octopus 11.26.0, the current release on 19 November 2026. Install the plugin in Claude Code:
Export the variable, then start a research run:
Before probe_discover dispatches its perspectives, Claude Octopus sends the question to Keenable. Each perspective gets up to five titles, URLs and 500-character excerpts, wrapped as untrusted content. They are leads, not verified claims: the perspectives still open and cite the pages, and the usual citation checks apply.

Behaviour

One lookup has a 15-second deadline and doesn’t retry or follow redirects. Python’s standard library makes the request, so there is nothing else to install. The key, if you set one, is kept out of model child processes and prompts.
Without a key, lookups run on the shared public tier at lower rate limits, shared by everyone behind your IP address. Export KEENABLE_API_KEY to use your organization’s limits. See API keys.

Gajae-Code

Use Keenable as the web search provider in Gajae-Code (gjc), the coding-agent harness, with providers.webSearch set to keenable. No key needed.

FrontierAgent

Set WEB_SEARCH_PROVIDER=keenable to send FrontierAgent’s web_search tool to Keenable, up to 50 results per query. Works without an API key.

Codex

Add Keenable to Codex over MCP or the Plugins Directory, and turn off Codex’s built-in web search so the agent calls Keenable.

CLI

Install the Keenable CLI, sign in once, add Keenable to Claude Code, Cursor and Codex over MCP, and search, fetch and run monitors from the terminal.