MFS: Unified Context for AI Agents

🔗 GitHub: https://github.com/zilliztech/mfs MFS, short for Multi-source File-like Search, is an open-source context harness that gives AI agents one searchable and browsable layer across scattered working data. Why MFS: ✦ Unified context layer — maps heterogeneous sources into one file-like URI tree ✦ Broad connector coverage — works with local files, GitHub, Notion, Slack, Google Drive, S3, BigQuery, Postgres, and more ✦ Search and Browse — combines high-recall discovery with precise evidence navigation ✦ Traceable retrieval — preserves links to original sources so agents can verify every conclusion ✦ Agent-native skills — provides mfs-ingest and mfs-find for Claude Code, Codex, and other agents ✦ Flexible integration — supports local or remote servers, SDKs, CLI, HTTP APIs, skills, MCP servers, and plugins In this video, we introduce the MFS architecture, supported data sources, and agent skills before walking through an end-to-end demo. We ingest evidence from Notion, BigQuery, Slack, local meeting notes, and GitHub, then ask an agent to reconstruct how a MemSearch threshold decision evolved. MFS distinguishes the shipped default from a later proposal, identifies a separate on-demand workflow, filters out irrelevant discussions, and cites the evidence behind each conclusion. ✨ If MFS is helpful to you, we would really appreciate a star on GitHub. We would also love to hear about your use cases, ideas, and feedback. Chapters: 00:00 What is MFS? 01:23 Architecture and the unified URI tree 01:52 Quick start and supported data sources 03:57 Demo setup: five sources for one real question 06:55 Ingest and index multiple sources 08:30 Cross-source retrieval with traceable evidence 10:16 Search and Browse 12:16 Build applications on MFS Find us on: Zilliz Website: https://zilliz.com Milvus Website: https://milvus.io Zilliz LinkedIn: https://www.linkedin.com/company/zilliz/ Milvus LinkedIn: https://linkedin.com/company/67143129 Milvus X: https://x.com/milvusio Milvus GitHub: https://github.com/milvus-io/milvus Try Zilliz Cloud: https://cloud.zilliz.com/signup