Xyne
AI-first Search & Answer Engine for work. Open-source alternative to Glean.
Last verified:
What is Xyne?
Xyne is an AI-first Search & Answer Engine for work that unifies fragmented information across your SaaS applications like Google Workspace, Atlassian suite, Slack, GitHub, and Zendesk. It securely indexes your data and builds a semantic relationship graph to provide unified search and accurate AI-powered answers with cited sources from your internal data.
Key features include Agentic RAG that grounds LLM responses in context from your apps, model-agnostic design allowing you to choose any LLM (AWS Bedrock, OpenAI, Ollama, Together AI, Fireworks AI, Gemini), self-hosted deployment anywhere (laptop, on-prem, cloud), permissions-aware access control that enforces your existing app permissions, and complete Google Workspace integration including Gmail, Calendar, Contacts, Drive, and attachments. Xyne does not train on your data or prompts and collects no telemetry.
Xyne is an open-source alternative to Glean, Gemini, and Microsoft Copilot, released under Apache 2.0 license. It is designed for individuals and organizations seeking a private, secure Work AI platform—.engineers looking for files and documentation, product teams asking about releases, support triaging issues, sales researching customer calls, ops finding policies, and anyone needing to discover the right people and information quickly.
The platform is built as an open, modular, extensible Work AI Platform with future roadmap including API for building internal AI applications, shareable AI Assistants with custom sources, and agents to automate multi-step workflows. The founding team has experience building large-scale projects like UPI, BHIM app, Juspay, and Hyperswitch.
Xyne pricing
Pricing model: Freemium
Xyne is completely free and open-source under Apache 2.0 license. There are no paid plans or subscription tiers. The software itself is free forever, but users must provide their own AI provider (AWS Bedrock, OpenAI, Ollama, Together AI, Fireworks AI, or Gemini) and bear the associated API costs. Deployment requires self-hosting on your own infrastructure (laptop, on-premises server, or cloud VM), so you pay for your own hosting resources.
Xyne pros
- Open-source under Apache 2.0 license
- Self-hosted - deploy on laptop, on-prem, or cloud
- Model-agnostic - works with any LLM provider
- Complete Google Workspace integration including Gmail attachments
- Permissions-aware - respects existing app access controls
- No training on your data or prompts
- No telemetry on usage
- Semantic graph maps relationships between knowledge, people, projects
- Agentic RAG for accurate answers with cited sources
- Multi-threaded data ingestion for high performance
- Atlasian suite ecosystem integration (not just isolated apps)
- Docker and Docker Compose deployment
- Supports multiple AI providers (AWS Bedrock, Ollama, OpenAI, Together AI, Fireworks AI, Gemini)
- Future-proof solution for bringing AI into work
- Extensible platform for building internal AI applications
Xyne cons
- Requires Docker and Docker Compose installation
- Chat functionality requires separate AI provider API keys
- Slack and Atlassian integrations still in development
- Complex setup requiring Google credentials and environment variables
- No official managed cloud hosting - self-hosting required
- Users must bring their own LLM provider and pay for API usage
- Initial model downloads can take 10-15 minutes
- PostgreSQL database required for deployment
Frequently asked questions about Xyne
What is Xyne?
Xyne is an AI-first Search & Answer Engine for work that connects to your applications like Google Workspace, Atlassian suite, Slack, and GitHub, securely indexes your data, and maps a relationship graph. It provides a Google + ChatGPT-like experience to find anything across all your apps and get up-to-date answers with cited sources from your internal data.
Is Xyne open-source?
Yes, Xyne is open-source under the Apache 2.0 license. It is positioned as an open-source alternative to Glean, Gemini, and Microsoft Copilot.
Where can I deploy Xyne?
Xyne is self-hosted and can be deployed anywhere - on your laptop, on-premises servers, or in the cloud. Deployment requires only Docker and Docker Compose, and can be run with a single docker-compose command.
Which AI models does Xyne support?
Xyne is model-agnostic and supports any LLM of your choice including AWS Bedrock, OpenAI (gpt-4o), Ollama (local models like DeepSeek), Together AI, Fireworks AI, and Google Gemini. You can even point it to local models via Ollama.
What applications does Xyne integrate with?
Xyne currently has complete Google Workspace integration including Drive, Gmail (with attachments), Calendar, and Contacts. Slack and Atlassian Suite integrations are next on the roadmap.
How does Xyne handle permissions and security?
Xyne is permissions-aware with live enforcement of all your apps' granular permissions, so users only see results based on what they have access to. It does not train on your data or prompts and collects no telemetry on usage.
Do I need to pay for Xyne?
Xyne itself is completely free and open-source under Apache 2.0 license with no paid plans. However, you need to provide your own AI provider and pay for their API usage separately. You also need to cover your own hosting costs.
How do I set up Google Workspace integration?
You need to generate Google credentials (GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET) by following Google's OAuth setup guide, then add them to your .env.default file along with other required environment variables like ENCRYPTION_KEY, JWT_SECRET, and GOOGLE_REDIRECT_URI.
What is Agentic RAG?
Agentic RAG is Xyne's implementation that grounds LLM responses in context from your apps. It provides accurate answers by retrieving relevant information from your indexed data and citing sources, with the maximum TPM exceeding for OpenAI's gpt-4o, so AWS Bedrock or Claude is recommended for best experience.
What is on Xyne's roadmap?
Beyond enterprise search with permissions-aware real-time RAG, Xyne's roadmap includes an API for building internal AI applications, shareable AI Assistants with custom sources, and agents to automate multi-step workflows. The goal is to build an open, modular, extensible Work AI Platform and Org Operating System (OrgOS).