Hybrid search, not a semantic lottery
Vector and full-text search are combined via reciprocal rank fusion and optionally re-ranked by a cross-encoder.
Connect SharePoint, Google Drive, Confluence, websites and files. Knowledgeworker Connect delivers reliable, source-backed answers inside ChatGPT, Claude, Cursor and your own applications.
Supplier contracts are retained for ten years after the contract ends, per the internal policy. Separate deletion periods apply to personal attachments.
Knowledgeworker Connect links your existing information sources to the AI tools your teams already use — without a new data silo and without bypassing the permissions of the source systems.
The core capabilities are not buried in long feature lists but made visible as an understandable data flow.
Vector and full-text search are combined via reciprocal rank fusion and optionally re-ranked by a cross-encoder.
Resolvable citations trace right down to the chunk of the original source.
Source ACLs are synced and enforced inside the database at retrieval time.
Partitions for departments and projects with their own extractors and access grants.
One read-only MCP server and a REST API with a generated TypeScript client.
Documents are synced incrementally, processed and made findable in a permission-true way. Start with one source and expand later.
Microsoft Graph delta, site and item ACLs.
Details → GDShared drives, native docs and sharing.
Details → CFSpaces, pages, attachments and restrictions.
Details → WWWSSRF-safe, rate-limited and robots-friendly.
Details → ↑PDF, Office, Markdown, text and OCR.
Details →Tenant isolation, source permissions and deletion propagation are part of the technical architecture — not just a promise on the website.
Your teams ask in familiar interfaces — answers stay bound to sources and permissions.
Technical documentation, guidelines and project knowledge as controlled read-only context.
Integrate search, fetch and answer into portals, service tools or industry-specific applications.
A standardized MCP endpoint for current AI clients and a REST API for custom integrations.
// .cursor/mcp.json { "mcpServers": { "knowledgeworker": { "url": "https://kwrag.chemmedia.ai/mcp", "headers": { "Authorization": "Bearer $KW_API_KEY" } } } }
const answer = await kw.answer({ query: "Which policy applies?", partition: "quality-management", citations: true });
The Free tier includes one connector and 1,000 queries per month.