SAG
A new SOTA for RAG — an original retrieval architecture and an open-source knowledge base for humans and agents.
SAG is an MCP server and knowledge base application built on a novel retrieval architecture that outperforms traditional RAG and GraphRAG approaches. It combines event-entity indexing with query-time dynamic hyperedges to deliver semantic retrieval and relational reasoning in a single system, achieving state-of-the-art results on multi-hop QA benchmarks. Developers can expose their knowledge bases to AI agents through the Knowledge MCP server, OpenAI-compatible endpoints, or the zleap-sag Python package. The system includes document ingestion, vector and multi-strategy search, source tracing, knowledge graph visualization, and cited agent conversations.
Key Features
Use Cases
- 01Exposing a personal or team knowledge base to Claude Code, Codex, or other MCP-compatible AI assistants
- 02Building multi-hop question-answering systems that need to traverse relationships across documents
- 03Creating cited AI responses grounded in specific document sources for compliance or verification
- 04Integrating document search and retrieval into existing applications via REST API or Python package
- 05Replacing traditional RAG pipelines with a single system that handles semantic and relational queries
- 06Exploring knowledge graphs of events and entities extracted from uploaded documents
Related MCP Servers
View moremarkitdown
Python tool for converting files and office documents to Markdown.
firecrawl
The context API to search, scrape, and interact with the web at scale. 🔥
prompts.chat
f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.
langflow
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
SAG — FAQ
What is SAG?+
SAG is both an MCP server and a complete knowledge base application built on a novel retrieval architecture that combines event-entity indexing with query-time dynamic hyperedges. It can be used as a standalone knowledge base or integrated into AI agents via MCP, OpenAI-compatible endpoints, or the zleap-sag Python package.
How do I install and connect SAG to my AI agent?+
Install the official @zleap-ai/sag-cli package globally with npm, then run 'sag agent connect codex' or 'sag agent connect claude-code' to automatically mount the MCP server. The CLI handles discovery, verification, and configuration without manual JSON editing.
Which AI clients work with SAG?+
SAG works with any MCP-compatible client including Claude Code and Codex. It also provides an OpenAI-compatible chat endpoint that works with any client supporting the OpenAI API format.
What API keys or prerequisites are required?+
SAG requires an OpenAI-compatible LLM endpoint and embedding model for document processing and search. Docker Desktop or Docker Engine with Compose v2 is needed for the self-hosted deployment; Node.js 20+ is required for the CLI; Python 3.11+ for local development.
Is SAG free to use?+
Yes, SAG is released under the MIT License and is free and open-source. You only pay for the LLM and embedding API calls to your chosen provider.
Can I use SAG without Docker?+
Yes, SAG offers desktop installers for macOS 15+ and Windows 10/11 that bundle the complete stack, or you can run local development with Python 3.11+ and Node.js 20+.
How do I install SAG?+
Open the source repository on GitHub and follow its README. SAG is a mcp server — MCP Agents Market links you directly to the official repo.
Is SAG free?+
SAG is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.