basic-memory

Search, read, and write notes in a persistent Markdown knowledge graph via MCP tools.

4.0k|282|Updated Dec 2, 2024
One-click install
npx skills add https://github.com/basicmachines-co/basic-memory --skill basic-memory-basicmachines-co
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: basic-memory
Source: https://github.com/basicmachines-co/basic-memory/tree/main/integrations/hermes/skill
Command: npx skills add https://github.com/basicmachines-co/basic-memory --skill basic-memory-basicmachines-co

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI assistants forget everything between sessions, forcing users to re-explain decisions, projects, and context. This Skill connects the assistant to a persistent Basic Memory knowledge graph so prior decisions, meetings, and insights can be searched, recalled, and captured across sessions. ## Core Features & Use Cases - Search before answering: Use bm_search and bm_read to recall prior decisions, project notes, and meeting outcomes before responding. - Capture knowledge: Use bm_write and bm_edit to save decisions, meeting notes, and insights as structured Markdown notes with observations and relations. - Multi-project routing: Use bm_projects, bm_workspaces, and workspace-qualified permalinks to read and write notes across local and cloud projects. - Use Case: A user asks "what did we decide about the auth strategy?" — the assistant runs bm_search, reads the matching decision note, and answers with the recorded context instead of guessing. ## Quick Start Ask the assistant to search Basic Memory for notes about a past decision or project before answering your question.

Frequently Asked Questions about basic-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I give my AI assistant persistent memory across sessions?▼

Connect the assistant to a Basic Memory knowledge graph through MCP. The bm_search, bm_read, and bm_write tools let it recall prior notes and capture new decisions as Markdown files that persist between sessions.

How do I search Basic Memory notes before answering a question?▼

Call bm_search with a query string and optional limit, then use bm_read on the returned identifier to fetch the full note. bm_search combines semantic and full-text matching to find relevant notes.

Should I use the bm CLI or the bm_* MCP tools?▼

Use the bm_* MCP tools for note operations. They route through a persistent MCP connection at roughly 0.1 seconds per call, while the CLI spawns a fresh Python process each time and bypasses automatic per-turn capture.

Can Basic Memory work with multiple projects or cloud workspaces?▼

Yes. Call bm_projects to list projects and bm_workspaces for cloud workspaces, then route read and write calls using a workspace-qualified project name like "personal/main" or a stable project UUID.

What note format does Basic Memory use for observations and relations?▼

Notes are Markdown with YAML frontmatter. Lines like "- [decision] ..." become observations, and WikiLink lines under a Relations heading become graph relations, making notes searchable and traversable.