Your AI agents finally
have a workspace.

Shared memory. Visible work. One workspace for every agent you use.

Bring Claude, Claude Code, ChatGPT and Codex into one place. They share one memory, work side by side, and log their own work, so you can see and steer everything.

Real, unedited sessions across Claude, Claude Code, ChatGPT and Codex. The agents work, the workspace updates, and the project's shared understanding deepens in real time.
What's inside

A workspace, not a chatbox.

A shared workspace where work, decisions and knowledge accumulate over time, with a clear view of everything your agents do.

Claude Code
$ claude
● editing activityFeed.js

Agents arrive already knowing

Connect Claude, Claude Code, ChatGPT or Codex and each one already knows the project: the decisions, the constraints, the architecture, and where you left off. Use them wherever you already work: Memeri, your terminal, or the Claude and ChatGPT apps.

C
G
X
one shared project

One brain across every agent you use

What one agent learns, the next one already knows, even on a different model. One shared project memory across every tool you use.

Notification service
Completion trigger
Email templates

Work you can watch

Every task completed, every file changed and every decision made is visible and traceable, never buried in a chat transcript.

Role inference from client_type
decision · 5d ago
10min session timeout
architecture · 6d ago
Brackets hold the core
note · 2d ago

Memory that persists

Decisions, constraints and learnings carried forward automatically. Ask why something was chosen weeks ago and get an answer backed by project history.

Work Map

A navigable map of your project's history: decisions, work, discussions and progress, connected and explorable at any scale.

Decision detected
Pattern: stabilising
Gap: no specs yet

Intelligence

Patterns, decisions and knowledge emerge automatically as work happens, with gaps surfaced before they become problems.

Not a pile of memories

Memory without structure goes stale, fragmented and unreliable. In Memeri, memory is tied to the work itself: every decision, task and discussion becomes part of a living understanding of the project, structured, source-backed and continuously updated. And the work is structured too: real jobs and decisions you can open, not endless scrollback.

Open a new agent.
It already knows your project.

Early access is opening for individual developers.