AI Developer Tools
PMB
Local-first memory for AI coding agents. Automatically stores and recalls project context via MCP, offline, with ~35ms recall.
What is PMB?
PMB is an open-source, local-first memory system for AI coding agents. It automatically stores and recalls project context (decisions, lessons, facts) using hybrid search (BM25 + dense vectors + entity graph) and hooks into agents like Claude Code, Cursor, and Codex via MCP. All data resides on your disk in a single SQLite file; no cloud, no telemetry.
PMB vs Similar AI Tools
| Pricing Model | Free | Free | Custom Pricing | Free, Paid |
| Free Credits | ||||
| Key Features |
|
|
|
|
| Pros |
|
|
|
|
| Cons |
|
|
|
|
| Best For |
|
|
|
|
How to use PMB?
- 1Install: `pip install pmb-ai`
- 2Connect your agent: `pmb connect claude-code` (supports Cursor, Codex, Zed, etc.)
- 3Work as usual – PMB records and recalls automatically.
- 4Open the dashboard: `pmb dashboard`
PMB Key Features
- Local-first: all data on your disk in SQLite + LanceDB
- Hybrid recall (BM25 + dense vectors + entity graph) with RRF fusion
- Auto-recall on every agent prompt (4–16 ms read)
- Sub-millisecond async writes (<1 ms, non-blocking)
- Lesson scoring: flags unused memories for pruning
- Live entity graph dashboard with timeline and map
- Works with multiple agents (Claude Code, Cursor, Codex, Zed, Windsurf)
- Fully offline, no API keys, no telemetry
- Apache 2.0 open source
PMB Use Cases
- Stop re-explaining your project to coding agents every session
- Switch between agents (Claude Code ↔ Cursor ↔ Codex) without losing context
- Maintain honest, self-cleaning memory that highlights helpful lessons
PMB Pricing & Free Credits
PMB currently operates on a Free model.
This tool is completely free to use
PMB Pros & Cons
Pros
- 100% local and offline
- Open source (Apache 2.0)
- Fast recall and writes
- Works with multiple MCP-aware agents
- Automatic memory hygiene (decay, archive, dedup)
- No API keys or cloud dependencies
Cons
- Requires Python environment (pip install)
- Limited to MCP-compatible agents
- Still relatively new, smaller community
What is PMB best for?
- Developers using AI coding agents who want persistent project memory
- Teams sharing context across multiple agents
- Anyone valuing data ownership and offline operation