Self-hosted alternatives to PMB
200 self-hosted apps like PMB, closest match first.
Top picks
- E

Engraphis
LaunchLocal-first, inspectable memory for coding agents: durable context across sessions and repositories, code-aware recall, bi-temporal history, MCP, and a self-hosted WebUI.
Open sourceSelf-hosted
- R
Roampal
LaunchOutcome-based memory for AI coding tools. Tracks which advice actually helped and learns from results. Works with Claude Code and OpenCode. 100% local, open source.
Open sourceSelf-hosted
- M
Marm Memory
LaunchLocal-first 3-in-1 AI memory layer & MCP server for Claude Code, Codex, Grok, Gemini, VS Code and Cursor. Fuses session history, codebase indexing & concept graphs in SQLite. Enables zero-cloud, privacy-first context & instant recall, suppo
Open sourceSelf-hosted
- G
GrayMatter
LaunchGrayMatter gives AI agents persistent memory. It ships as a single binary of about 10 MB that you install without Docker, a database server, config files or a cloud account. It works as a general-purpose MCP server with Claude Code, Cursor,
Open sourceSelf-hosted
- M
Mnemory
LaunchA self-hosted, secure, feature-rich memory system for AI agents and assistants. Provides intelligent fact extraction and deduplication, with an artifact store for detailed content.
Open sourceSelf-hosted
- B

Basic Memory
LaunchOne knowledge base for you, your AI tools, and your team. Your chats, docs, decisions, and research, connected in a living knowledge graph. Plain Markdown you control. Open source, MCP-native, works with every AI.
Open sourceSelf-hosted
More like PMB
Metronix Memory
Self-hosted AI agent memory — MCP memory server with durable recall for agents
About PMB
PMB
Local-first memory for your AI coding agent
PMB is a local-first memory for AI coding agents such as Claude Code, Cursor, Codex and Zed. It stores decisions, lessons and project facts in one SQLite file on your disk, with vectors in LanceDB beside it. It works offline, with no API keys and no cloud, and the site says there is no telemetry. Agents connect through MCP, and hooks inject the right memory before the model replies. Recall is hybrid, combining BM25 keyword search, dense vectors and an entity graph, fused and ranked, with optional reranking. The site says recall takes about 35 milliseconds, and writes return in under a millisecond because embedding and vector insert run on a background thread. Lessons are ranked by how often they are followed. A dashboard shows memory as a graph, a timeline of decisions, lessons and commits, and the site includes a map view. PMB is installed with pip install pmb-ai and is licensed under Apache 2.0.
- Price
- Open source
- Category
- AI
- Website
- pmbai.dev




