Skip to content
TribalMind

Introduction

TribalMind is a shared memory layer for AI development agents. It lets your AI coding assistants — Claude Code, Cursor, Windsurf, GitHub Copilot, and others — remember, recall, and share knowledge across sessions, projects, and team members.

The problem

Section titled “The problem”

Every time you start a new AI agent session, your agent starts from scratch. It doesn’t know:

  • The gotchas your teammate discovered last week
  • Why that config flag is set the way it is
  • The workaround for that flaky dependency
  • How your modules connect or why things are built a certain way

You end up repeating context, re-explaining decisions, and watching agents make the same mistakes.

How TribalMind solves it

Section titled “How TribalMind solves it”

TribalMind gives your agents a persistent, semantic memory with pluggable providers — Backboard (default) or Mem0.

Terminal window
# Store knowledge in natural language
tribal remember "numpy 1.26 breaks on Python 3.13 — pin to <1.26"


# Find relevant knowledge by meaning, not keywords
tribal recall "numpy compatibility issues"

Memories are:

  • Persistent — they survive across sessions and restarts
  • Semantic — search by meaning, not exact keywords
  • Shared — your whole team (and all their agents) can access the same knowledge
  • Cross-repo — search across all projects in your account with --all

How it works

Section titled “How it works”
  1. tribal init — connects your project to Backboard and creates an assistant (a scoped memory namespace)
  2. tribal remember — stores knowledge; an LLM parses your natural language into structured entries
  3. tribal recall — retrieves relevant memories using semantic search
  4. tribal setup-agents — generates config files so your AI agents automatically use TribalMind

Key features

Section titled “Key features”
FeatureDescription
Agent-nativeWorks with Claude Code, Cursor, Windsurf, Copilot — no plugins needed
Semantic searchVector-powered retrieval finds knowledge by meaning
Cross-repo search--all flag searches memories across every project in your account
Structured memoryLLM auto-parses entries into category, subject, and content — including workflows
Pluggable providersChoose Backboard (default) or Mem0 as your memory backend
Team knowledgeShared memory across all team members and their agents
Dashboard UIWeb interface for browsing and managing memories
JSON outputEvery command supports --json for programmatic use
Stdin pipingPipe knowledge in and out via standard streams

Next steps

Section titled “Next steps”