The problem
AI coding agents are useful, but the memory they build can include architecture choices, security assumptions, customer constraints, and internal codebase context. For many teams, that memory belongs inside the company boundary.
Self-hosted AI memory
Self-hosted memory for AI coding agents
your infra
Palace gives engineering teams persistent agent recall without sending decisions, conventions, fixes, or project history to an external memory cloud.
Runs on your infrastructure, not Palace's cloud.
Local ONNX embeddings: no third-party embedding API required.
Ed25519-signed licenses verify offline for air-gapped operation.
How Palace works
Developers keep using their normal MCP-capable tools. Palace Server runs beside your code and stores searchable memory in your Postgres deployment, with local embeddings and tenant-scoped access controls.
Who it is for
Small engineering teams, startups, and regulated organizations that want the speed of agent memory without making cloud-hosted memory a new data exposure surface.