Repolode
Submit
codesstar

codesstar/loci

A local, Markdown-and-JSON memory and data layer shared across AI coding agents like Claude Code and Codex, viewable and editable by the user.

View on GitHub
Stars
499
Forks
15
Open issues
5
Last commit
14 days ago

What is it?

Loci gives several AI agents one shared personal memory. Preferences, tasks, schedules, contacts, project context and saved links are kept in a local "brain" directory as Markdown, JSON and attachments, so an agent connected to it can pick up the same context instead of starting from scratch each time.

Agents are pointed at the brain through a short entry file, LOCI.md, plus a full rules manual, LOCI-RULES.md, which they read when a memory operation first comes up. Data is loaded on demand rather than all at startup. Project-specific memory stays in the project repository, with the brain holding only an index.

Users can browse and edit the files directly or use a local Dashboard built on Node.js. No external database or cloud memory service is required. Whether an agent actually follows the rules depends on its file and command tools and how it is configured, since installing the files does not force model compliance.

Key features

  • 01Shared memory across Claude Code, Codex, Hermes, Qwen, Doubao, Qoder and WorkBuddy
  • 02Data stored locally as Markdown, JSON and attachments that can be edited by hand
  • 03Local Dashboard for viewing and changing saved records
  • 04Tasks, schedules, contacts and project context kept in one brain directory
  • 05On-demand loading of data to avoid reading everything at startup

Use cases

  • Keeping preferences and background consistent when switching between AI coding agents
  • Tracking project status and technical decisions while keeping the index in one place
  • Managing personal tasks, meetings and saved links through an agent
  • Keeping a local, portable backup of AI-accumulated notes

Quick start

  1. 1

    Send the README's install prompt to your AI agent (Claude Code, Codex, etc.) to install Loci

  2. 2

    If Loci already exists, send the connect prompt to the new tool to link it to your existing brain

  3. 3

    Ask your agent to open the Loci Dashboard using your existing brain (keep the local service running)

Things to know

  • Requires Node.js 18+, Git, and local file or command access from the agent.
  • The Dashboard runs as a local Node.js service that must stay running.
  • Rule files guide the model but cannot force it to follow them.
  • Models or third-party services used by your agent may be paid.

Topics

#ai-memory#claude-code#local-first#markdown#ai-agent#personal-knowledge#mcp-alternative#ai#memory#open-source

FAQ

What is loci?
A local, Markdown-and-JSON memory and data layer shared across AI coding agents like Claude Code and Codex, viewable and editable by the user. Loci gives several AI agents one shared personal memory. Preferences, tasks, schedules, contacts, project context and saved links are kept in a local "brain" directory as Markdown, JSON and attachments, so an agent connected to it can pick up the same context instead of starting from scratch each time.
Is loci free to use? What license does it have?
loci is open source under the MIT license. Check the license terms before using it in commercial products.
Is loci still maintained?
Actively maintained. The last commit to loci was on Oct 8, 2026.
What can I use loci for?
Keeping preferences and background consistent when switching between AI coding agents Tracking project status and technical decisions while keeping the index in one place Managing personal tasks, meetings and saved links through an agent Keeping a local, portable backup of AI-accumulated notes
How do I install loci?
Send the README's install prompt to your AI agent (Claude Code, Codex, etc.) to install Loci If Loci already exists, send the connect prompt to the new tool to link it to your existing brain Ask your agent to open the Loci Dashboard using your existing brain (keep the local service running)
Where can I get loci?
The source code and installation instructions are on GitHub: https://github.com/codesstar/loci

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

Related projects