Dimple documentation

Dimple stores information for AI agents and finds it again later. Use the SDK in an application, the CLI in a terminal, or the MCP server from an MCP client.

Choose a path

If you want to	Start here
Store and search memories in an application	Quickstart
Use Dimple from a terminal	CLI guide
Let an agent call Dimple	MCP server
Search a codebase	Index a codebase
Understand how Dimple organizes and retrieves memories	Architecture
Configure storage, embeddings, or search	Configuration reference

How Dimple works

A write stores a memory. An index run stores source files. Dimple keeps both in a local database by default and creates embeddings for semantic search. A query runs semantic search and keyword search, then combines the ranked lists for the agent.

The topic tree groups memories as the collection grows. The artifact graph records relationships between memories, such as support, conflict, and supersession. Background jobs maintain both structures.

Start with local storage

The default setup uses local SQLite files and a local embedding model. To change the embedding provider or use a remote libSQL database, see the configuration reference.

Next steps

• Follow the quickstart to write and search your first memory.
• Connect an MCP client to give an agent access to Dimple.
• Read the architecture to learn how storage, retrieval, and background jobs work.