01
decision2m
Use PostgreSQL advisory locks
Source: Nelson · architecture.md
PROJECT MEMORY · AVAILABLE NOW
Your repo
should remember.
MemHub gives every developer and coding agent the decisions, constraints, fixes, and discoveries that already shaped the project.
Create a shared memory
Local-first by default. Hosted or self-managed. No external embedding API required.
acme/payments-api
synchronizedcontext/main
memories 38
pinned rules 7
contributors 13
BRIEFING FOR NEW SESSION
What changed since your last pull
02
pattern18m
API errors follow RFC 9457
Source: Maya · pull request #184
03
bugfix1h
Refresh tokens rotate atomically
Source: Codex CLI · auth/session.ts
Claude Code
Cursor
Codex CLI
THE MECHANISM
Context enters once. Everyone inherits it.
01
Commit locally
A memory is written to SQLite and the durable outbox before the network is involved.
02
Resolve its project
Trusted credentials attach organization, repository, author, and version provenance.
03
Brief every session
Authorized humans and agents pull only the context added since their last sequence.
Connect the tools already in the repository.
No new editor. No proprietary chat interface. No background polling loop holding your work hostage.
repository terminal
$ npx @memhub/mcp connect --ide codex✓ project resolved · context ready
Host it your way.
Keep local data regardless. Run the shared layer yourself or let MemHub operate it.
Team
$29 / month
- 5 project contexts
- 5 team seats
- 10,000 synchronized memories
- Pinned rules and audit activity
Questions without the sales pitch.
Does MemHub replace agentmemory?
No. agentmemory remains the fast local engine. MemHub adds project synchronization, access control, audit, and a shared team portal.
What happens when the shared service is offline?
Nothing interrupts the work. Writes commit to the local SQLite outbox and synchronize when the service returns.
Can we self-host it?
Yes. The repository includes Docker and Easypanel configuration for PostgreSQL 16 and pgvector.
Do we need an external embedding API?
No. Start with full-text search, connect Ollama locally, or configure a supported remote provider later.
THE NEXT SESSION STARTS HERE