Fabric provides an evolving memory layer that instantly indexes an entire codebase into a structural knowledge graph, capturing functions, dependencies, and decisions so AI agents can understand code context and impact of changes. The shared, vendor‑agnostic knowledge persists across tools like Claude, Cursor, and Codex, allowing sessions to compound and agents to start with existing context rather than from scratch. Built‑in governance features detect drift, resolve contradictions, and ensure auditable, trustworthy answers.
Funding
Funding not disclosed
Founders
Product
Problem
AI coding assistants often operate without shared, persistent knowledge of a codebase, requiring each tool to re‑index files and leading to inconsistent context, missed dependencies, and frequent hallucinations about code behavior.
Solution
Fabric provides a local‑first memory layer that indexes an entire codebase in seconds, constructing a structural knowledge graph of functions, dependencies, and design decisions. This curated graph is stored centrally and made available to multiple AI agents—such as Claude, Cursor, and Codex—so each can instantly retrieve accurate, auditable context without re‑indexing. The platform continuously evolves: every session adds new information, while built‑in governance, drift detection, and contradiction resolution keep the graph consistent and trustworthy. Agents receive verified recall and zero‑hallucination responses, enabling reliable code understanding, impact analysis, and automated refactoring across tools.
Target Audience
Primary users are development teams that employ multiple AI coding assistants and need a unified, trustworthy code intelligence layer to improve productivity and reduce errors.
Features
- Instant full‑codebase indexing that maps functions, dependencies, and decisions into a typed knowledge graph
- Persistent, evolving memory that compounds automatically across sessions and agents
- Built‑in governance, drift detection, and contradiction‑resolution engine with full audit trails
- Cross‑provider portability via MCP tools, supporting Claude, Cursor, Codex, and custom agents
- Verbatim recall and structural grounding to eliminate hallucinations and ensure token‑efficient context
- Integration connectors for GitHub, Obsidian, and Slack to sync external content into the graph