SWIRL 5 is generally available today. It is the most significant release we have ever shipped, and it sharpens SWIRL into one idea: the private knowledge layer that makes your chosen LLM and agents better - federating search and RAG across everything you already have, with no second copy of your data.
Here is what's new, and why each piece earns its place.
A home dashboard built around your knowledge
SWIRL 5 opens on a dashboard that organizes knowledge into workspaces and topics. Recent files from every connected source - OneDrive, Box, iManage, ServiceNow, and more - surface right there, one click to add to a workspace. It is the difference between a search box and a place your team actually works.
Canonical answers your organization approves
Federated search finds documents; SWIRL 5 decides which one is authoritative. Teams pin the canonical result for a query - the approved policy, the governing clause, the right precedent - and every later search, by a person or an agent over MCP, returns the endorsed answer instead of the model's best guess.
A three-pass relevancy pipeline, all local
Relevance now runs in three passes: keyword and BM25 (quotes and exact terms honored first), embedding re-ranking with hybrid fusion, then a cross-encoder that reads the query and document together to score real relevance. Both models run locally - nothing is sent over the wire - and there is no vector database to build or secure.
A first-class MCP server for your agents
SWIRL 5 is headless and API-first, with a first-class Model Context Protocol server. Any agent - Claude, Copilot, ChatGPT, or your own - calls SWIRL as its governed knowledge layer and gets ranked, permissioned, organization-approved answers. No copy of your data leaves your tenant. It is the safest way to give agents enterprise reach.
A hallucination warning when answers aren't grounded
Every generated answer is checked against the sources it cited. When a claim isn't supported by what was actually retrieved, SWIRL flags it - so your team knows when to verify, and nothing ungrounded slips through unnoticed.
Still the same promise
All of it holds to the principle SWIRL was built on: your data stays where it lives, permissions are enforced live, and there is no index or vector store to govern. Private by architecture, the answer instead of a guess, and the lowest-risk on-ramp to enterprise AI.
SWIRL 5 is available now. To see it against your own stack, request preview access - it includes a 30-minute guided stand-up on a slice of your systems.