Rubi

Rubi is a context manager and knowledge graph that reduces token costs and keeps sessions short for AI coding assistants.

Rubi is a context manager and knowledge graph for developers using AI coding assistants. It shortens conversational sessions by compacting early, saving a handoff state, and restoring it immediately to lower context re-reading costs. In addition, it builds an import map and a temporal knowledge graph of bugs, decisions, gotchas, and concepts directly in a local markdown folder. The engine searches this graph using BM25, Personalized PageRank, and Reciprocal Rank Fusion within a strict token budget. It runs locally as a standard Model Context Protocol (MCP) server, integrating with tools like Claude Code, Codex, Gemini CLI, Cursor, and GitHub Copilot.

Rubi at a glance

Best for: Software engineer
Runs on: VS Code, CLI
Pricing: Free
Released: October 2026
Category: AI coding tool
Website: github.com/yuribedran/rubi

Key features

Compacts sessions early to prevent context bloat
Saves and restores handoff states during compaction
Maps code imports and centrality using PageRank
Maintains a temporal knowledge graph of bugs, decisions, and concepts
Combines BM25, Personalized PageRank, and Reciprocal Rank Fusion for search
Stores all project knowledge as plain markdown files compatible with Obsidian
Operates locally as an MCP server with zero dependencies
Provides a token X-ray tool to analyze transcript costs

Use it for

Lowering token costs and context re-read expenses during long AI coding sessions
Retrieving project-specific architectural decisions and bugs without manual searching
Navigating codebases with hierarchical map summaries instead of reading whole repositories
Tracking superseded technical decisions and related git commits over time

Alternatives to Rubi

CodeSight — Reduces AI token usage significantly and maps projects instantly.
CodeGraph — It eliminates file crawling, reducing AI agent tool calls by 88% and cutting costs.
Oracle — Provides a single entry point for multiple pro AI agents and session management.
jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
Codebase Memory MCP — High-speed indexing and persistent knowledge graphs for AI coding agents.
ContextMCP — Maintains fresh, accurate context for AI agents using AST-aware chunking.
AI Memory — Shared memory for coding agents that works across machines and team members.
Memorax Code — Carry project architecture and verified engineering decisions across coding sessions.
CodeGraphContext — Turns codebases into interactive knowledge graphs for AI assistants.
Dosu — It enables a shared knowledge cache so agents avoid redundant data gathering.
ContextStream — It maintains project context and history to reduce agent errors.
Context7 — Provides up-to-date documentation for AI agents and code editors.

All Rubi alternatives