Giving AI coding agents memory that survives a restart

Description and use

AI agents retain context across restarts, auto‑detect project workspace, and share memory via a lightweight Rust server.

Problem

AI coding agents lose context after a restart, forcing repeated explanations of decisions, conventions, and project history.

Outcome

Agents retain context across restarts, automatically detect project workspace, and share knowledge via a lightweight Rust binary that runs on small VMs or Raspberry Pi, enabling air‑gapped deployments.

Workflow

• 1. Install palace-rs core on local machine.
• 2. Agent automatically gains local persistent memory via palace_* MCP tools.
• 3. For team use, run palace remote set to point to shared Palace Server.
• 4. Run palace remote on to enable shared memory access.
• 5. Agents now read/write to shared memory, preserving decisions, diaries, and permissions.
• 6. Team members benefit from shared knowledge graph and session context without workflow changes.

Evidence

palace-rs is the open source core (MIT licensed). Install it, and your agent gets a local, persistent memory.

Source

For a team, Palace Server adds the shared layer on top: shared knowledge graph, team diaries, workspace permissions.

Source