Repolode
Submit
aayoawoyemi

aayoawoyemi/Ori-Mnemos

Local-first persistent memory for AI agents, stored as markdown files with a knowledge graph, retrieval, and an MCP server.

Open-source alternative to:Mem0

View on GitHubWebsite
Stars
335
Forks
32
Open issues
1
Last commit
last month

What is it?

Ori Mnemos gives AI agents memory that persists across sessions, clients, and machines. Notes are plain markdown files on disk, wiki-links act as graph edges, and a local SQLite index handles semantic and keyword search, so no database server, cloud service, or API key is needed for core retrieval.

Retrieval combines semantic embeddings, BM25 keyword matching, personalized PageRank, and an associative warmth signal, then reranks results using learned Q-values and co-occurrence edges. Memory decays over time using ACT-R style equations, with separate decay rates for identity, knowledge, and operational notes. The project also includes a recursive exploration command that decomposes a question into sub-questions and traverses the graph.

Ori ships as a CLI, a library, and an MCP server with 16 tools. It includes bridge commands that connect Claude Code, Hermes Agent, OpenCode, Cursor, and Codex to a shared vault.

Key features

  • 01Stores memory as markdown files with wiki-links forming a knowledge graph.
  • 02Fuses semantic, BM25, PageRank, and warmth signals for retrieval.
  • 03Decays notes over time with ACT-R-style forgetting and spreading activation.
  • 04Exposes 16 MCP tools and a CLI for use by any MCP-capable agent.
  • 05Runs embeddings locally, with no API key required for core features.

Use cases

  • Giving a coding agent persistent project context across sessions.
  • Sharing one memory vault between Claude Code, Cursor, and other MCP clients.
  • Keeping agent memory in a git-versioned folder that you can edit in Obsidian.
  • Running a self-hosted agent on a VPS with memory stored on the filesystem.

Quick start

  1. 1

    Scaffold a vault

    npx ori init
  2. 2

    Build the index

    npx ori index build
  3. 3

    Run a navigated retrieval query

    npx ori explore "…"

Things to know

  • The learned retrieval data (Q-values, LinUCB arms, retrieval history) is not derived from markdown, so it must be exported before rebuilding the index.
  • Benchmark results are self-reported, and the README notes that some comparisons are against older or differently measured systems.

In collections

Long-term memory for AI agentsOpen-source AI agent frameworks & appsMCP serversPrivacy & local-first toolsDatabases & SQL toolsStatic site generators & documentation tools

Topics

#agent-memory#mcp#knowledge-graph#local-first#sqlite#semantic-search#markdown#ai-agent#ai-agents#ai-memory

FAQ

What is Ori-Mnemos?
Local-first persistent memory for AI agents, stored as markdown files with a knowledge graph, retrieval, and an MCP server. Ori Mnemos gives AI agents memory that persists across sessions, clients, and machines. Notes are plain markdown files on disk, wiki-links act as graph edges, and a local SQLite index handles semantic and keyword search, so no database server, cloud service, or API key is needed for core retrieval.
Is Ori-Mnemos free to use? What license does it have?
Ori-Mnemos is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is Ori-Mnemos still maintained?
Updated within 6 months. The last commit to Ori-Mnemos was on Oct 1, 2026.
What can I use Ori-Mnemos for?
Giving a coding agent persistent project context across sessions. Sharing one memory vault between Claude Code, Cursor, and other MCP clients. Keeping agent memory in a git-versioned folder that you can edit in Obsidian. Running a self-hosted agent on a VPS with memory stored on the filesystem.
How do I install Ori-Mnemos?
Scaffold a vault: npx ori init Build the index: npx ori index build Run a navigated retrieval query: npx ori explore "…"
Where can I get Ori-Mnemos?
The source code and installation instructions are on GitHub: https://github.com/aayoawoyemi/Ori-Mnemos

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects