Meet BrunoSend a request, commit the collection. No account, no cloud.
Search Stack
Managed search API your AI agent can run end-to-end
10 followers
Managed search API your AI agent can run end-to-end
10 followers
Full-text, vector, geo, and image search over one JSON API — with typeahead, a drop-in autocomplete widget, and official TypeScript/.NET clients. What makes it different: the entire index is controllable via MCP (read and write, not just querying), so an AI agent can create lists, add records, and tune search for you. And built-in evaluations let you prove a relevance change is actually better before you ship it. Lucene-backed, no cluster to babysit.
How does the MCP read/write access work in practice when an agent makes changes — is there version history or some way to roll back if an automated tweak tanks relevance on a key query?
How does the MCP write access handle conflicts if my agent and a teammate are both updating records at the same time, and is there any audit trail to see what changed?
How does the MCP write access actually get scoped in practice? Like, can I lock an agent to only mutate a specific index or field so it can't accidentally rewrite production data?
how does the MCP write access handle permissions when multiple agents could be tuning the same index at the same time
the MCP write access is a really thoughtful move, letting agents shape the index instead of just querying it. paired with built-in evals, that's the kind of craft that makes shipping relevance changes feel safe.