Local and self-hosted tools for agent memory
Agents and assistants that can run on your own machine or server, keeping data local. Of the 105 tools on our memory for agents page, 20 fit this narrower job. By pricing: 19 open source, 1 source-available. The most-starred open repos here are Headroom (75,629 stars), MemPalace (59,622 stars), Nanobot (49,096 stars). Most come from the memory & RAG (11), personal agents (6), productivity (1) categories.
Updated
Headroom
Library, proxy and MCP server that compresses tool outputs, logs, files and RAG chunks locally before they reach the LLM, cutting tokens for coding agents.
Added
MemPalace
Local-first AI memory system that stores conversations verbatim with a pluggable backend and serves recall to agents without API calls.
Added
Nanobot
Ultra-lightweight, self-hosted personal agent framework in Python with a web UI, tools, memory and MCP support.
Added
Supermemory
Memory and context engine for agents with an API, app and MCP server, which can also run fully locally.
Added
llmware
Framework for building private, local RAG and agent applications with small specialized models on AI PCs, edge and self-hosted setups.
Added
EverOS
Local-first, portable memory layer for AI agents stored as Markdown.
Added
xerj
A local-first search engine and MCP server for agents: one Rust binary indexes code, docs, logs, CSV/JSONL and PDFs for full-text, vector and hybrid search plus agent memory, behind an Elasticsearch-compatible API.
Added
OpenKnowledge
OpenKnowledge is a free, local-first markdown editor and knowledge base built for working alongside AI agents. It looks and feels like a Notion-style WYSIWYG editor, with callouts, tabs, Mermaid diagrams and media blocks, but every page stays a plain markdown file on your disk. Coding agents such as Claude Code, Codex, Cursor, Gemini CLI and dozens of others can read and write the same files through its integrations, so specs, team wikis and LLM wikis stay in one place that both people and agents keep up to date. It ships as a desktop app plus a web UI and CLI.
Added
LongMemory
Local-first memory engine for agents that keeps durable, time-aware memories in SQLite and serves them to Claude, Copilot, Codex and others over MCP, CLI or HTTP.
Added
Agent Sandbox (Kubernetes)
A Kubernetes project that adds a Sandbox resource for running isolated, stateful agent workloads such as code execution and browsers, with warm pools and fast startup on your own cluster.
Added
Basic Memory
Persistent memory for Claude, Codex and Cursor stored as local Markdown knowledge, available open source or as a hosted cloud workspace.
Added
Mercury Agent
Self-hosted personal AI agent with a written personality, local memory, permission-gated tools and token budgets, reachable via CLI or Telegram.
Added
Webcmd
Browser layer for AI agents that learns how each website works as agents use it and saves that as local memory to make later runs faster and cheaper.
Added
Memmy
A local-first personal AI agent with a self-evolving long-term memory that its desktop app, CLI and your other AI agents can all share.
Added
Jarvis
Private, local-first desktop AI assistant with voice and text chat, personal memory, dictation and MCP tools.
Added
OpenLore
A self-hosted knowledge repository for AI agents written in Go, where agents and teams share docs, notes and findings over SSH and MCP with scoped access.
Added
projectmem
A local-first memory layer for coding agents that records issues, attempts, fixes and decisions and warns your agent before it repeats a past mistake.
Added
Graphicle
Graphicle, from KarLex AI, is a single local Rust program that gives Claude Code, Cursor, Codex and other MCP agents a code graph and memory, returning excerpts with file and line references from your code, docs and past sessions, with no account or telemetry.
Added
Dodaios
Open-source desktop agent OS for Windows, macOS and Linux (formerly Personal Jarvis): talk to it and it runs your coding CLIs side by side, your browser and your apps, with offline Whisper speech, a local wiki memory and support for Ollama, Claude or ChatGPT subscriptions.
Added
TecAdRise Bot
TecAdRise Bot is a free, MIT-licensed Windows desktop app (Electron and Cursor SDK) for running a local team of AI agents, each with its own model, memory, skills, MCP servers, workspace and schedule.
Added
GitHub repos (20)
| Repo | GitHub stars | Licence | Last push |
|---|---|---|---|
| 75,629 | Apache License 2.0 | 2026-10-21 | |
| 59,622 | MIT | 2026-10-20 | |
| 49,096 | MIT | 2026-10-21 | |
| 31,281 | MIT | 2026-10-18 | |
| 14,814 | Apache License 2.0 | 2026-10-01 | |
| 13,451 | Apache License 2.0 | 2026-10-21 | |
| 11,507 | — | 2026-10-21 | |
| 4,646 | — | — | |
| 4,541 | — | 2026-10-18 | |
| 4,328 | — | 2026-10-20 | |
| 4,263 | — | 2026-10-20 | |
| 3,433 | — | 2026-10-21 | |
| 2,725 | — | 2026-10-20 | |
| 2,125 | — | 2026-10-19 | |
| 2,097 | — | 2026-10-19 |
Top 15 of 20 by stars. Stars and licence come from each repo, checked when the listing was last verified.
FAQ
What are the best AI tools for memory for agents focused on local and self-hosted?
We list 20. The first three by GitHub stars and featured status are Headroom: Library, proxy and MCP server that compresses tool outputs, logs, files and RAG chunks locally before they reach the LLM, cutting tokens for coding agents.; MemPalace: Local-first AI memory system that stores conversations verbatim with a pluggable backend and serves recall to agents without API calls.; Nanobot: Ultra-lightweight, self-hosted personal agent framework in Python with a web UI, tools, memory and MCP support.
Are there free or open-source options?
Yes: 19 open source, 1 source-available. Pricing is as each maker states it.
More from memory for agents
Tools are grouped by what each says it does. Missing one? Submit it free.