SWIRL is an open-source federated AI search and RAG platform that answers questions across your company’s existing apps and data sources without moving the data. Developed by swirlai under the Apache-2.0 license, the community edition has about 3,050 stars on GitHub and is actively developed in 2026. Instead of building yet another index, SWIRL queries sources live, merges the results and uses large language models to re-rank them and compose an answer with citations.
SWIRL ships with connectors for more than 100 sources, including tools like Microsoft 365, Google Workspace, Slack, GitHub, BigQuery, Snowflake and generic SQL, web and REST endpoints. When a user asks a question, SWIRL dispatches it to the selected sources in parallel, normalises and deduplicates the returns, applies LLM-based relevance ranking, and can generate a retrieval-augmented answer that links back to the original items. Because nothing is copied into a new index, permissions and freshness stay with the source systems, which simplifies enterprise governance considerably.
SWIRL also provides the connective tissue many internal AI assistants are built on: its RAG pipeline can feed retrieved, permission-aware content into chat interfaces, and its connector library covers the long tail of enterprise systems, from document stores to ticketing tools. Deployment is container-based, with an admin console for configuring sources and monitoring queries. For organisations wary of centralising sensitive content into a third-party index, SWIRL’s query-time federation model is its defining architectural choice, and the open-source edition is fully functional for evaluation and smaller deployments.
The SWIRL community edition in this repository is free and open source under Apache-2.0. The company also sells commercial SWIRL offerings with enterprise connectors, support and deployment options, making the overall product freemium.
IT and knowledge teams at companies whose information is scattered across many SaaS tools, and developers building enterprise search or internal RAG assistants.