Memory — Slate docs

Every agent you use shares Slate's memory. It is stored as docs you can read.

Slate is an MCP server. A tool that speaks MCP (Claude Code, Codex, Cursor, GitHub Copilot, Antigravity, OpenCode, Gemini, Cline, Grok Build, Kiro, Qwen Code, Droid, Hermes Agent, Goose, Kilo Code, Auggie) starts a session with the project's memory index, searches with recall and saves with remember. One workspace, many tools.
A watcher fills it in. It reads the conversation files your agents already write and stores each session as a doc. It also reads the memory folders Claude Code, Gemini CLI and Codex keep for themselves, for the folders you added as projects.
Memory is your docs. Sessions, decisions and other memories live in the same local SQLite workspace as everything else. Fix a wrong doc and the next recall reads the correction.

In this section#

Doc	What it covers
Session Capture	The watcher that captures AI sessions as docs.
Recall & Remember	Search and citing, with links to the session brief, lessons, saving and tidying.
Decisions & Memories	Logged choices with rationale, and durable knowledge ranked for recall.
Memory Graph	Memories on the files they cite, coloured by use.

The tools every AI client calls are in Slate as an MCP Server.

recall reads the workspace on this machine that owns the working folder, and never leaves the machine. What you put in a shared workspace is readable by colleagues and their AI through a scope of its own, recall with scope: "workspaces", which asks the cloud.