MindStone Agent documentation

MindStone Agent is documented as a family of continuity layers and product components.

The important idea is simple: memory is not a feature. It is a layered continuity system. Identity, logs, handoff, structured memory, recall, context management, and communication each do one job; together, they help agents resume work with better judgment instead of reconstructing everything from scratch.

Choose your path

Claude Code

Use MindStone for Claude Code if you want persistent identity, memory, checkpointing, handoff, and recall inside Claude Code.

git clone https://github.com/MindStone-Agent/mindstone-for-claude-code.git
cd mindstone-for-claude-code/orchestrator
./bootstrap.sh

Pi

Use MindStone for Pi if you want a Pi-native MindStone Agent continuity layer.

curl -fsSL https://raw.githubusercontent.com/MindStone-Agent/mindstone-for-pi/main/install.sh | bash

Then in Pi:

/ms-init
/ms-onboard
/ms-status
/ms-recall-status

Agent communication

Use Synapse when separate persistent agents or humans need a shared communication plane for review, coordination, and handoff.

Synapse is also the human-in-the-loop path: people can participate directly in the same communication channels agents use to coordinate, review, interrupt, approve, or redirect work.

Local AI command center

Use Cortex when you want a self-hostable dashboard for a trusted local AI stack: system/GPU telemetry, installed local models, model discovery and local Ollama pulls, service health, and local tool controls.

Coming-soon paths

MindStone Agent Harness — the native harness product.
MindStone Agent Packs — Dockerized, ready-to-run role agents.
MindStone for Codex — Codex continuity layer.