Browse documentation
MCP tools
The eleven tools this server advertises, generated from what tools/list actually reports.
Every tool below takes the same parameters as its REST endpoint and
returns the same body, and each carries an outputSchema, so a client can
type the result rather than parsing a string. Connect a client from
the client list.
web_search Search the agent-first web index. Returns metadata-only hits (url, host, snippet, score, trust and graph signals) — never the page body; fetch that with get_document when a snippet is not enough.
get_document Fetch the full cleaned text of a passage by its id (from a web_search hit), with its metadata.
similar More-like-this: given a passage id from a web_search hit, return its semantic neighbours.
why_not Coverage transparency: is this URL in the index, and if not, what happened to it?
browse Browse without a query: the most-recent (or top-host-rank) passages, optionally within a vertical.
related Coverage Graph: the connected neighbourhood of a passage — 'follow the thread'.
corroborate Coverage Graph: the independent-source view of a passage's story — how many DISTINCT hosts carry it, and which passages, grouped by source and ranked by host authority.
authorities Coverage Graph: the top passages about a topic (or entity), ranked by graph centrality rather than by match to a query.
dossier Coverage Graph: a one-hop brief on an entity — the passages that mention it, its top source hosts, and the entities it co-occurs with.
path Coverage Graph: the shortest chain of edges linking two passages — connect-the-dots retrieval.
assemble_context GraphRAG-as-a-service: instead of a link list, a ready-to-read context pack — the corroborated, story-deduped, trust-ranked passages for a question, packed to a token budget, each carrying the reason it was included.