PMB

Local-first memory for your AI coding agent, stored in a single SQLite file on your disk.

Visit

PMB provides AI coding agents like Claude Code, Cursor, and Codex with persistent memory. It stores decisions, lessons, and project facts in a local SQLite file, ensuring offline operation without API keys or cloud reliance. Recall is fast, around 35ms.

---

How to use PMB?

Install PMB with 'pip install pmb-ai'. Connect your agent using 'pmb connect <agent_name>'. PMB automatically records and recalls information during your coding sessions. Explore your memory through the local dashboard or terminal commands.

---

PMB 's Core Features

Local-first memory storage in SQLite

Offline operation, no API keys or cloud required

Hybrid recall (BM25, dense vectors, entity graph)

Automatic memory injection before agent reasoning

Memory hygiene: decay, compaction, and self-tending

Agent-agnostic via MCP protocol

Open-source (Apache 2.0)

---

PMB 's Use Cases

Giving AI coding agents persistent memory

Maintaining context across multiple coding sessions

Switching between AI coding tools without losing context

Building a personal knowledge base for AI development

Ensuring data privacy with local-only storage

Improving AI agent recall accuracy and speed

Automating context management for AI coding

---

PMB 's Pricing

---

PMB 's FAQ

Does my code or data ever leave my machine?

How is this different from RAG or a vector database?

Will it slow my agent down?

Which agents and operating systems are supported?

What if a memory is wrong or unhelpful?

Is it really free?

---

Most impacted jobs

Software Engineer

AI Developer

Machine Learning Engineer

Data Scientist

Programmer

---

PMB Youtube Reviews

---

PMB 's Tags

#AI Memory#Coding Agent#Local-first#Offline#SQLite#Vector Database#RAG#Open Source

---

PMB 's Alternatives

NexaLibre

Let your AI agent build and deploy your website or full-stack app on secure, managed hosting.

Memmy

A local-first AI agent with a self-evolving memory foundation for seamless context sharing across tools.

gstack joins your meeting

Bring Garry Tan's gstack specialists into your Google Meet as real voice bots with 3D avatars.

Velane

Integration lifecycle infrastructure for AI agents, enabling autonomous workflow creation, testing, and deployment.

FluentDB

A native macOS database client with an AI co-pilot that writes SQL alongside you.

Second Brain

Give Claude, ChatGPT, and Cursor permanent memory, self-hosted and free forever.

OpenWorker

AI agent that runs on your computer, works in your tools, and turns requests into finished deliverables.

Inkling

Inkling is an open-weights, multimodal, Mixture-of-Experts model with controllable reasoning effort, available for fine-tuning.

BaseRT

The fastest LLM runtime for Apple Silicon, enabling local model execution on your device.

NexaLibre

Let your AI agent build and deploy your website or full-stack app on secure, managed hosting.

Memmy

A local-first AI agent with a self-evolving memory foundation for seamless context sharing across tools.

gstack joins your meeting

Bring Garry Tan's gstack specialists into your Google Meet as real voice bots with 3D avatars.

Velane

Integration lifecycle infrastructure for AI agents, enabling autonomous workflow creation, testing, and deployment.

FluentDB

A native macOS database client with an AI co-pilot that writes SQL alongside you.

Second Brain

Give Claude, ChatGPT, and Cursor permanent memory, self-hosted and free forever.

OpenWorker

AI agent that runs on your computer, works in your tools, and turns requests into finished deliverables.

Inkling

Inkling is an open-weights, multimodal, Mixture-of-Experts model with controllable reasoning effort, available for fine-tuning.

BaseRT

The fastest LLM runtime for Apple Silicon, enabling local model execution on your device.