Skip to content
Dhanasvi
Sign inExplore
musecl-memory logo

musecl-memory

Verified

Git-powered portable memory that keeps AI agents' knowledge safe across machines.

Autonomous AgentsAgent Frameworks 6Open source
View on GitHub
Updated 2026-06-15
musecl-memory GitHub repository

What is musecl-memory?

musecl-memory treats accumulated agent knowledge like source code. Memory files live in a git repository while derived vector indexes stay local and are rebuilt on each machine. The design avoids large binary blobs and keeps everything auditable and under user control.

On every sync the tool compares files by content hash, copies only changed sources, scans for secrets, then commits and pushes. Pull operations bring fresh sources to any machine where the agent immediately rebuilds its own indexes from the shared files.

It suits developers and teams running agents across laptops, servers, or ephemeral VMs who need memory to survive hardware changes without vendor lock-in or privacy trade-offs.

Capabilities

sync agent memory via files
version memory with git
store data in markdown
replace vector db persistence
zero-dependency operation

What you can build with musecl-memory

Surviving hardware failure

When a disk dies or a VM is recycled, the agent can restore its full knowledge base on replacement hardware by cloning the git repo and rebuilding indexes.

Moving between development machines

Developers can push memory updates from a laptop and pull them onto a workstation, ensuring both environments share the same corrections and learned patterns.

Running agents in short-lived environments

CI runners or temporary containers can pull the latest memory sources, rebuild indexes on the fly, and discard the instance without losing any accumulated intelligence.

Install musecl-memory

Quick start
Traditional Software                    AI Agent Memory
======================                  ======================
package.json         <--->              MEMORY.md (source)
package-lock.json    <--->              patterns.json (source)
node_modules/        <--->              vectordb/, *.sqlite (derived)
npm install          <--->              agent memory index --force
git push/pull        <--->              sync.sh push/pull
  1. 1Clone the musecl-memory repository to a sync folder on your machine.
  2. 2Point your agent workspaces at the local memory directory.
  3. 3Run the push script after the agent writes new memory files.
  4. 4Run the pull script on any other machine to receive updates.
  5. 5Rebuild the vector index locally after each pull.

musecl-memory: pros & cons

Pros

  • +Memory stays fully portable and private via standard git workflows
  • +No binary files or embedding mismatches are ever synced
  • +Works with any agent that can read Markdown and JSON sources
  • +Secret scanning prevents accidental credential leaks

Cons

  • –Requires manual or scripted rebuild of indexes after every sync
  • –No built-in conflict resolution for simultaneous edits on multiple machines
  • –Depends on users maintaining their own git hosting and backup strategy
Did you find this helpful?

Frequently asked questions

No, only the source Markdown and JSON files are committed; vector indexes are always rebuilt locally.

User reviews

Verified reviews from the community shape this listing's rating.

Loading reviews…

Sign in to review

Promote musecl-memory

Add this badge to your website, or share the tool.

DFeatured on Dhanasvimusecl-memory 0