Overview
Knowledge Plane is a shared memory layer designed to streamline workflows and facilitate communication by allowing teams to work in their preferred AI toolslike Cursor, Claude, and otherswithout creating new silos.
It aims to consolidate a team's code, docs, and chats into a comprehensive and up-to-date source of truth that is auditable across existing tools. The system automatically extracts and cleans data from ingested documents or synced GitHub repos.
Knowledge Plane combines Graph capabilities that model relationships with Vector Embeddings for understanding semantics. It is designed to plug into AI agents like Claude and Cursor through the Model Context Protocol (MCP), with no custom integrations needed.
The platform auto-ingests existing documentation and codes to build knowledge cards and also offers self-hosting options for those who prefer to keep proprietary context within their infrastructure.
Knowledge Plane isnt a simple vector store but combines graph memory with vector embeddings to allow AI agents to reason about dependencies, ownership, and timelines.
Additionally, it features workspace isolation, full audit logs, and scoped API keys for security concerns.
Key Features
- Knowledge Cards — Atomic Facts Extracted From Docs, Not Raw File Storage. Higher Accuracy Under Tight Context Windows • Skills — Scheduled Autonomous Jobs That Detect Drift And Update Facts Automatically (hourly/daily). This Is A Real Differentiator, Most Tools Don't Have This • Graph Memory With Typed Edges (depends_on, Decided_by, Owned_by) — Agents Can Traverse Chains Of Decisions Like A Senior Team Member Would • Hybrid Retrieval: Graph Traversal + Vector Search Combined • Auto-ingestion From Google Drive, Github Repos, Pdfs, Wikis, Tickets
Releases
Top alternatives
-
Emily Richardson🙏 48 karmaDec 30, 2024@MemoriPyMemoripy is a GAMECHANGER. I see why it has so many stars on GitHub- we use it for all of our agents to ensure they are context aware! Great job 👏
-
I'm using MemoryPlugin since a year. It is brilliant how can it improve the work with LLM-s. With any LLM-s. Asad did a lot of improvement, he is open for discussion about enhancements.
