PMB

What is PMB?

PMB is a local-first memory layer for AI coding agents that keeps decisions, lessons, and project facts in a single SQLite file on a developer’s machine. It plugs into MCP-aware tools like Claude Code, Cursor, Codex, Zed, and Windsurf so each agent shares the same long-term context across sessions. By recalling relevant snippets in roughly 35 milliseconds and writing updates asynchronously, it gives AI assistants persistent project awareness without any cloud, telemetry, or API keys.

Key Features:

• Local-first persistent memory: Stores all events in SQLite with vectors in LanceDB on disk, so agents read and write project memory without data leaving the machine.
• Automatic hybrid recall: Classifies every message in under a millisecond, then uses BM25, dense vectors, and an entity graph with reciprocal-rank fusion to fetch useful context before the model answers, with reported recall@10 of 94.6 percent on the LoCoMo dataset.
• Low-latency reads and async writes: Recall typically lands in tens of milliseconds, while writes return in under 1 millisecond as embeddings and vector inserts run on a background thread.
• Map and Timeline dashboard: A local web app shows a live entity graph and git-style timeline of decisions, lessons, commits, and facts so developers can inspect and clean up memory.
• Honest lesson scoring: Tracks whether the agent actually follows each rule, starring helpful lessons and flagging ignored ones as dead so teams avoid bloated or misleading memory.
• MCP-native integration and BYO model: Runs as a Model Context Protocol server over stdio, wired with commands like pmb connect claude-code, and can use hosted models or a fully local Ollama instance for summaries and graph extraction.

Who is Using PMB?

• Backend and full-stack engineers: Use it to give coding agents durable knowledge of APIs, edge cases, and architectural decisions.
• Tooling and platform teams: Standardize a shared memory layer across Claude Code, Cursor, Codex, Zed, and similar tools for internal repos.
• Security-conscious engineering orgs: Adopt AI coding help while keeping source code and project metadata fully local and inspectable.
• Indie developers and OSS maintainers: Keep long-running project context in one workspace that survives agent restarts and editor changes.
• Uncommon Use Cases: Data scientists tying experiment notes and findings to MCP-aware notebook agents; instructors running lab setups where student coding agents remember project history across sessions.

Pricing:

• Open source package: Free to install and use under the Apache 2.0 license, with no paid tier or seat-based pricing listed.

Disclaimer: Please note that pricing information may not be up to date. For the most accurate and current pricing details, refer to the official PMB website.

What Makes PMB Unique?

PMB combines a strict local-first design with automatic hybrid recall that runs before the model reasons and a feedback loop that scores which memories truly change agent behavior. Paired with a visual entity graph, git-like timeline, and one inspectable SQLite file wired natively to MCP-aware coding agents, it offers unusually transparent, measurable AI memory on a developer’s own disk.

How We Rated It:

• Accuracy and Reliability: 4.5/5
• Ease of Use: 3.7/5
• Functionality and Features: 4.2/5
• Performance and Speed: 4.8/5
• Customization and Flexibility: 4.0/5
• Data Privacy and Security: 4.9/5
• Support and Resources: 3.8/5
• Cost-Efficiency: 5.0/5
• Integration Capabilities: 3.6/5
• Overall Score: 4.3/5

Local Memory Layer For Coding Agents:

PMB suits developers who already rely on MCP-aware coding agents and want those agents to remember project details without sending code to the cloud. Teams that care about privacy, latency, and seeing which memories actually help will get strong value from its dashboard and honest metrics. It is less suited to non-technical users or non-MCP tools, but for engineering groups comfortable with the terminal, it offers a fast, transparent memory layer that fits naturally into the development stack.