About Ditto
Ditto gives AI agents a shared workspace for conversations, files, projects, and tasks. Its memory system connects information to subjects such as people, decisions, and preferences, helping agents retrieve relevant context across threads.
Through MCP, builders can make that memory available to agents in tools such as Claude, Codex, and Cursor.
Memory research on Bittensor
Ditto runs Bittensor Subnet 118 (SN118), where miners develop programs that help agents store information and recall it when needed. These programs are called agent-memory harnesses.
Validators evaluate submissions with DittoBench, measuring memory recall and tool-calling accuracy against fresh datasets. Developers can use the official starter kit to build and practice before submitting a harness. The mining guide explains the workflow, requirements, and scoring.
Web search in conversations
Ditto also supports searches for current information within a conversation. It can summarize relevant results and provide source links for further reading. See its web search documentation for details.
