Reads along
It picks up your goals, decisions, questions and taste as you work. All on your machine.
Recallstack remembers your goals, what you decided and why, what's still open, and how you like things done, across every agent you use.
Agents can read a codebase, rebuild its history and redo a change in minutes. What still gets lost is the thinking around it: what you're going for, what you decided and why, what you ruled out, and how you like things done.
Today that lives in your head and in old chat threads. Every new session, and every switch between Codex and Claude, starts with you explaining it again.
Each agent keeps its own notes too, like CLAUDE.md, AGENTS.md and Cursor rules. They drift apart and none of them follow you. Recallstack sits above all of them.
Picking up from Codex. We're starting fresh, doing the website first, and the homepage is next. I'll keep the copy plain. Want me to start on it?
It picks up your goals, decisions, questions and taste as you work. All on your machine.
Open another agent and it starts from where you left off. No catching it up.
If a plan goes against something you decided, the agent stops and asks you first.
Recallstack hooks into each agent and picks out goals, decisions, questions and taste from what you both said. It runs on your machine, with the agent you already have.
Each agent records a decision, a new question or a correction the moment it sees one, without asking you.
Your own words beat anything inferred, and a correction beats everything.
Every record links back to the exact words it came from.
Confirm, edit or delete anything, or mark it as a one-off.
Full conversations stay on your machine. Only the records are kept.
curl -fsSL https://github.com/o-clan/recallstack-cli/releases/latest/download/install.sh | bashWorks with Codex, Claude Code, Cursor, Copilot and Antigravity.