wayformFAQGet early access What is shared memory for coding agents?
Shared memory is one record of your team’s decisions, and the reasons behind them, that every teammate’s coding agent can read. With Wayform, a decision made in one Claude Code session reaches a teammate’s fresh Cursor or Codex session automatically. Nobody pastes context, and nobody re-argues a settled call.
Coding agents remember per person, per tool. Claude Code’s memory stays on one machine; Cursor rules and AGENTS.md hold what someone wrote down ahead of time. A team needs the decisions it makes as it works, and those usually live in chat scrollback.
How it works in Wayform
- Record the decision and its “because”. Ask your agent in plain language: “Remember we chose GitHub login because we don’t want a second auth system.” Plans also record their decisions when they ship.
- It lands in your space’s ledger. An append-only record, attributed to whoever wrote it, optionally mirrored to a private GitHub repo.
- Every session starts briefed. A compact briefing arrives when a session starts, and relevant pointers arrive with each prompt. Agents search memory before re-deciding something.
- Correct it in the open. When the team changes course, a new decision supersedes the old one. Review replacements and fix facts in the app’s Memory view.
An example
Maya decides in Claude Code on Friday that invitations reuse GitHub identity. On Monday, Chris asks Codex to add team invitations. Codex’s brief already includes Maya’s decision and reason, so it extends the existing login flow instead of proposing a new account system. If it tried, the guard would bring the decision back to Chris first.
What shared memory is not
- Not a transcript store. Wayform keeps settled decisions and durable context, not every message.
- Not a replacement for instruction files. Keep
CLAUDE.md,AGENTS.md, and Cursor rules for stable repo conventions. How they fit together. - Not a guarantee. Agents still make mistakes; the guard and plan check catch some of them, not all.
When you don’t need it
If you work alone in one tool and your conventions fit in an instruction file, that file is enough. Shared memory pays off when several people, or several agents, make decisions the others need to respect.
Set it up
- Quickstart: install the app and sign in with GitHub.
- Memory for Claude Code and memory for Codex.
- A teammate handoff across agents.
- Your data & access: where memory is stored and who can read it.
Questions
Does Claude Code’s memory work across a team?
Claude Code’s memory and instruction files are tied to one machine or one repo. Wayform adds a shared record of decisions and their reasons that every teammate’s Claude Code, Cursor, and Codex session reads through MCP.
Where is the shared memory stored?
In your space’s append-only ledger on Wayform’s hosted service, optionally mirrored to a private GitHub repo you own. Your chats, code, and build worktrees stay on your machine.
How does an agent get the team’s memory?
A compact briefing arrives when a session starts, and relevant pointers arrive with each prompt. Agents also search memory before re-deciding something.
What happens when the team changes its mind?
Record the new decision as a replacement for the old one. You review replacements in the app’s Memory view, and the old decision stays in the history.