Local-first AI memory for coding agents. Recall context instantly
Visit WebsiteFreeVisit Website
Tracked since2026
0 reviews trackedThe Bottom Line
Entry price
Free, no paid tier
Biggest pro
Fully local and open-source (Apache-2.0) — no data leaves your machine, no API keys needed
Biggest con
Requires initial setup and configuration for each agent (though simplified with pmb connect commands)
TL;DR - PMB
- Open-source, local-first memory for AI coding agents that automatically records and recalls context without cloud dependencies
- Injects relevant lessons, decisions, and project overview before the agent reasons, using sub-millisecond hybrid recall (BM25 + dense vectors + entity graph)
- Integrates with multiple agents (Claude Code, Cursor, Codex, etc.) via MCP and provides a local web dashboard to explore memory as a graph or timeline
Pricing: Free forever
Best for: Individuals & startups
What is PMB?
PMB is an open-source, local-first memory system designed for AI coding agents. It automatically records and recalls context from your coding sessions, injecting relevant memories (lessons, decisions, project overview) before the agent responds — without requiring additional API calls or cloud services. The tool uses a hybrid recall system combining BM25, dense vectors, and entity graphs to surface the most relevant information in milliseconds. PMB integrates seamlessly with popular AI coding agents like advanced models, Cursor, and Zed via the Model Context Protocol (MCP), and comes with a local web dashboard for exploring your memory as an interactive graph or timeline. All data stays on your machine in SQLite and LanceDB, and the system is fully offline-capable.
Pros & Cons
Pros
- Fully local and open-source (Apache-2.0) — no data leaves your machine, no API keys needed
- Works across multiple AI coding agents (advanced models, Cursor, Zed) without losing context
- Fast hybrid recall (minimal latency) with no LLM call on the read path, keeping latency minimal
Cons
- Requires initial setup and configuration for each agent (though simplified with pmb connect commands)
- Focused on AI coding agents; may not be designed for general AI assistant memory (e.g., chat bots outside coding)
Key Features
Automatic memory injection before the model thinks (pre-read) and journaling after actions (post-write)Hybrid recall with BM25, dense vector embeddings, entity graph, and optional reranker fused via Reciprocal-Rank-FusionSub-millisecond async writes: returns instantly to agent, embeds and vector storage happen in background threadMemory scoring: tracks whether each lesson is followed by the agent; flags ignored lessons as deadInteractive local dashboard with entity graph map, timeline (git-graph style), and overviewNo LLM call on the read path; optional summarization and graph extraction can run on local Ollama modelsMCP integration via stdio — no network, ports, or tokens requiredSupports multiple projects and agents sharing the same workspace
Pricing
Free
PMB is completely free to use with no hidden costs.
Reviews
$99Free with your review
Write a reviewReview PMB, get a free AI guide
Share your experience and we will send you Improve Your Thinking Patterns Using ChatGPT, free.
Best PMB Alternatives
Top alternatives based on features, pricing, and user needs.
FirecrawlFreemium
Turn websites into LLM-ready data for your AI applications.
WeavableFreemium
Give your AI agents persistent memory and context for more effective, long-running tasks.
LangChainFreemium
Build LLM-powered applications
Zep MemoryFreemium
Build personalized, fast, and reliable AI agents with comprehensive, evolving memory.
IgnitionRAGFreemium
An administrable AI backend for documents and agents, designed for multimodal RAG in production.
pumaDBFreemium
AI agent memory, simplified. No database admin needed
LlamaIndexFreemium
Data framework for LLM applications
Still deciding?
Most buyers shortlist 2 or 3 tools before committing. Pull a side-by-side comparison or browse the full alternatives shortlist below.
Explore More
PMB FAQ
How does PMB recall context from previous coding sessions without adding latency?
PMB uses a hybrid recall system that combines BM25, dense vectors, and entity graphs to surface relevant memories in milliseconds. Because the read path does not require an LLM call, the tool injects context instantly before the agent responds.
Which teams benefit most from adopting PMB for their development workflow?
Teams building with AI coding agents such as advanced models, Cursor, or Zed benefit most from PMB. Developers who need persistent, local-first memory across sessions without relying on cloud services will find the tool especially useful for maintaining project context and decisions.
Can PMB be used with general AI assistants like chat bots outside of coding?
PMB is focused on AI coding agents and may not be designed for general AI assistant memory such as chat bots outside coding. Its architecture is optimized for recording and recalling context from coding sessions specifically.
How is PMB priced and does it require any paid subscription?
PMB is free to use with no paid plan required. The tool is released under an open-source Apache-2.0 license, so there are no subscription fees or hidden costs.
How does PMB compare to Zep Memory for developers building with AI agents?
Unlike Zep Memory, PMB is fully local-first and open-source, keeping all data on your machine in SQLite and LanceDB without needing API keys. PMB also avoids additional API calls on the read path to keep recall latency minimal, whereas Zep Memory may rely on cloud-based retrieval.
What are the main trade-offs when using PMB with multiple coding agents?
PMB requires initial setup and configuration for each agent, though this is simplified with pmb connect commands. The tool is primarily focused on AI coding agents and may not serve as a general memory solution for non-coding AI assistants.
Does PMB integrate directly with Cursor and Zed without needing cloud services?
Yes, PMB integrates seamlessly with AI coding agents like advanced models, Cursor, and Zed via the Model Context Protocol (MCP). All integrations run locally with no cloud dependency because the system is fully offline-capable.
Which kind of memory does PMB automatically record during a coding session?
PMB automatically records lessons, decisions, and project overviews from your coding sessions. These memories are stored in a local web dashboard that visualizes them as an interactive graph or timeline for easy exploration.
Source: pmbai.dev