An MCP-ready memory server for AI agents

Squish provides memory operations for agent workflows that need to remember, recall, inspect, and manage durable project context through MCP and CLI surfaces.

Memory operations for agent workflows

Use the runtime to remember useful context, search prior knowledge, inspect retained memories, and retrieve relevant project state when a session starts.

Local-first by default

Start with a local runtime and SQLite-backed storage, then choose managed sync or shared infrastructure only when your team needs it.

7 tools for memory management

Squish exposes remember, recall, forget, link, context, stats, and inspect as MCP tools. Each handles a specific memory operation, from storing new observations to auditing what the system retains.

Frequently Asked Questions

What MCP tools does Squish provide?

7 tools: remember, recall, forget, link, context, stats, and inspect. Each handles a specific memory operation.

How do I add Squish to my MCP client?

Add the Squish MCP server config to your client's settings. The server runs locally via stdio transport.

Can I use Squish with multiple MCP clients?

Yes. Each client connects to the same local Squish runtime. Cloud sync shares memory across devices.

Read the setup docs

Squish documentation