Repository navigation
[Feature] Keenable as a web search provider that works without a key #4067
ilya-bogin-keenable
started this conversation in
Feature requests
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
The websearch skill now lets users pick a provider (Serper or You.com Eco, #3974). Both need an API key, so the skill stays off until a user signs up somewhere and pastes a key into
/mcp.Proposal: one more
webSearchProvidervalue,keenable, that works without a key. Serper stays the default, and nothing changes for existing users unless they pick it in/settings.skills/websearch/src/websearch/websearch.py: a_fetch_keenablenext to_fetch_eco. It POSTs the query tohttps://api.keenable.ai/v1/search/publicwith anX-Keenable-Title: Prime Agentheader and mapsresults[](title,url,snippet) onto the existing Serper result format, with each snippet capped.crates/pa-core/src/settings/types.rs(WebSearchProvider) and the skill gate incrates/pa-core/src/packages/manager.rs:keenableturns the skill on with no stored key. Keyless use allows 10 requests/s and 1000/hour per IP; a key slot can follow the Eco pattern later if anyone hits that.Keenable is currently first on the Artificial Analysis Search API index (83.0 vs Exa 80.3): https://artificialanalysis.ai/agents/search-api
If this fits, I can send the PR with tests mirroring the Eco ones once you invite it, or you can build it in-house.
Disclosure: I work at Keenable.
All reactions