agent-memory
Plain-Markdown long-term memory shared by Claude Code and Codex
Emerged finds new products worth a look, daily — what’s new today
If your assistant's memory is a folder of .dai files you can grep and carry between Claude, GPT and local models, you are no longer locked into one vendor's memory store.
DaiDocs treats AI memory as a file format rather than a service. Each .dai file is UTF-8 text in three parts: a YAML header, a JSON block holding what the model understood, and the original text. The analogy is JPEG: an open format any model reads. One command, npx daidocs setup, detects Claude Desktop, Claude Code, Cursor and Windsurf and wires in an MCP server plus an auto-save hook. On LongMemEval-S the maker reports 83% with GPT-4o and 92% with Claude Fable 5 while reading a tenth of the tokens. Apache-2.0, 26 stars, by Amin Rigi's Kerneta in London. A hosted version is on a waitlist.
More like this
Fraise
A single-binary memory database for AI agents with a two-verb query language
VoiceMem
Open streaming memory for real-time voice agents, split into facts and feelings
Strata
Shared memory for Claude Code and Codex sessions, with an LLM judge at the door
Aura
An agent appliance you host, with one memory database per person
FastRecall
Hosted conversation memory that follows you from one AI model to the next