Skip to main content
web_search backend ToolUniverse’s web_search tool can send its queries to Keenable: pass backend="keenable" in the call. ToolUniverse builds AI scientist systems from 2,700+ tools, and Keenable is one of its opt-in hosted search backends. It runs without an API key. The default auto chain doesn’t use Keenable, so nothing changes until you select it. → The Keenable backend in the ToolUniverse repository

Install

The Keenable backend ships in ToolUniverse 1.5.9 and later. ToolUniverse recommends installing with uv:

Search with Keenable

Load the web search tools, then call web_search with backend set to "keenable":
The response has status and a data object. data["results"] is the ranked result list, where each result has title, url, snippet and rank, and data["backend_used"] says which engine answered. Each snippet is Keenable’s short page summary, or the first 500 characters of the page text when there is none. web_api_documentation_search, the second tool in the same category, takes backend="keenable" too.

Arguments

If Keenable fails or returns nothing, ToolUniverse falls back to its auto chain, and data["provider_errors"] holds Keenable’s error. With the Keenable backend, the query goes to https://api.keenable.ai, so don’t send sensitive or patient-identifying information. See Data privacy.
If KEENABLE_API_KEY isn’t set, the backend calls /v1/search/public on the shared public tier at lower rate limits. Set KEENABLE_API_KEY to use your organization’s limits and remove the hourly cap. See API keys.

RL and eval environments

Use Keenable as the search backend in SkillOpt OfficeQA rollouts and Halo GRPO search environments: one config value, opt-in, never a default.

Spider

Search the web from the Spider Rust crawler with KeenableProvider, behind the search_keenable feature. Works without an API key.

AgentField

Run AgentField’s deep research example on Keenable web search with SEARCH_PROVIDER=keenable. No account or API key needed.

Tool calling

Give an LLM web search and page fetch as tools. Pass the SDK’s TOOLS to any OpenAI-compatible chat completions client and run each call with run_tool_call.