Company

Built in the open.
Grown from real need.

From a single open source commit to enterprise AI infrastructure running in law firms, federal agencies, and SaaS platforms - the SWIRL story.

2023
July 2023
v2.1
Community Edition - First Public Release
SWIRL releases under Apache 2.0. Federated search across multiple sources in a single query - with AI-powered re-ranking and result merging from day one.
Release notes →
August 2023
v2.5
RAG Pipeline and AI Summarization
Retrieval-Augmented Generation lands in the open source build. SWIRL becomes the retrieval layer for AI assistants, not just a search tool.
Release notes →
October - November 2023
v3.0
Microsoft 365 Integration
SWIRL reaches into the heart of the enterprise. Real-time federated search across M365 - SharePoint, Teams, Outlook, OneDrive - alongside any other source, simultaneously. No indexing. Permissions enforced by the source.
Microsoft 365 connector
Real-time permissioned retrieval
Simultaneous multi-source queries
Release notes →
December 2023
v3.1
Confidence-Based Scoring
Results ranked by confidence, not just relevance. Expanded connector library. SWIRL begins to understand the difference between a good match and a trustworthy one.
Release notes →
2024
February 2024
v3.2
Secure AI Mode
Enterprise hardening, data-at-rest protection, and secure AI mode for regulated environments. SWIRL earns its place in air-gapped and compliance-sensitive deployments.
Release notes →
October 2024
v3.8
Sub-Second Federated Query
Major performance release. Sub-second federated query across 10+ simultaneous sources. SWIRL scales to production workloads without architectural changes.
10+ concurrent source queries
Sub-second response times
Elastic horizontal scaling
Release notes →
November 2024
v3.9
MCP Server - SWIRL Enters the Agent Era
The Model Context Protocol changes how AI agents reach external knowledge. SWIRL ships a production-ready MCP server - one connection, 100+ enterprise sources, for any agent that speaks MCP. Claude, Copilot, and others gain enterprise knowledge access overnight.
MCP Server - production ready
Claude, Copilot, and MCP client support
No data pipeline required
Release notes →
2025
January 2025
Enterprise 4.0
AI Search + AI Assistant
The Enterprise line launches. Integrated RAG with citation UI, AI Assistant alongside federated search, and a co-pilot experience built for teams who need answers - not just results.
Integrated AI Assistant
Citation UI with source attribution
Enterprise deployment on Azure, AWS, GCP
Release notes →
April 2025
Enterprise 4.1
Smarter Citations and AI Fallback
Traceable citations with source verification. AI provider fallback so enterprise deployments never go dark. JSON Config API for programmatic control.
Release notes →
June 2025
Enterprise 4.2
Google Workspace Integration
Unified search across both Microsoft and Google. For the first time, a single SWIRL query reaches M365 and Google Workspace simultaneously - Gmail, Drive, Docs, and all of M365 in one result set.
Release notes →
September 2025
Enterprise 4.3
Deep Linking, Ratings, Japanese Support
Deep linking to source documents. User rating and feedback loop. Full Japanese language support - SWIRL goes multilingual for enterprise deployments in Japan.
Release notes →
November 2025
Enterprise 4.4
Document AI, Connect All, PII Redaction
Document-focused AI analysis. Connect All mode for rapid multi-source deployment. Automatic PII redaction for regulated industries.
Document-level AI analysis
Automatic PII detection and redaction
Connect All rapid deployment
Release notes →
December 2025
Enterprise 4.4.1
PDF Export, OAuth2, Stability
PDF export from AI Assistant sessions. OAuth2 in the Assistant UI. Stability improvements across high-volume deployments.
Release notes →
2026 · Released
2026
Enterprise 5.0
Zero-copy AI search
The question was never just "how do you find it?" It was "which version does your organization actually trust, and did you have to copy everything to find it?" SWIRL 5 answers both. Every search runs live as the signed-in user with permissions enforced at the source; results from every system are re-scored on one scale and grouped by version; and the one your organization approved leads the answer and is cited by it.
No copy of your content, no mirror of your permissions
Version clustering and the version that governs
Pinned results - organizationally approved answers
Semantic cache for instant repeat queries
Release notes →

SWIRL 5 is here.

Zero-copy AI search for enterprises that can't centralize their data. See it run on your own systems, and bring the auditor's questions.

Talk to Us