{"repo":"nikita-rulenko/Helixir","free":true,"listed":false,"github":"https://github.com/nikita-rulenko/Helixir","clone":"git clone https://github.com/nikita-rulenko/Helixir.git","description":"LLM reasoning outer weights memory layer to sync any agents you have","language":"Rust","stars":86,"topics":["ai","ai-memory","helixdb","mcp","rust","helixir","deep-learning","research","agentic-ai","reasoning"],"license":"MIT","category":"mcp-servers","readme_excerpt":"Helixir Persistent, governed, reasoning-aware memory for AI agents. One graph for Codex, Claude Code, Cursor, and every MCP-compatible client. Quick Start · What it is · How it works · Documentation --- Helixir is the memory layer an agent keeps when the model, editor, and session change. It extracts durable facts from conversations, preserves authorship, connects facts with typed reasoning edges, and recalls both an answer and the path that supports it. It is built for teams as well as individual agents. Permanent graph-backed RBAC keeps memories inside their groups, while explicit dedup federations let trusted groups share consensus without leaking knowledge across boundaries. Quick Start The portable installer selects the signed artifact for the current machine and opens the guided onboarding flow: Or install the native package first: Then converge the database, models, MCP clients, RBAC, and optional admin UI: Onboarding detects Codex, Claude Code, and Cursor, registers helixir-local , installs the canonical Agent Skill, and verifies a real embedding request. The default local path provisions mandatory NLI plus Ollama and nomic-embed-text ; an explicit remote embedding endpoint is also supported. After onboarding, restart the agent client once and ask it to recall Helixir memory. Global administrators can open the control plane at http://127.0.0.1:6971 . The APT repository setup, signing-key fingerprint, headless flags, three HelixDB topology choices, source builds, upgra","default_branch":null,"files":null,"tree":[],"storefront":"/r/nikita-rulenko","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/nikita-rulenko/Helixir/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}