Relay carries what your business decides straight into your developers' Claude — instantly.
Not the big decisions — those get announced. It's the quiet context: why you do it this way, the constraint you mentioned in passing.

Your Claude remembers you. It has no idea what the rest of your team just figured out.
What one person works out, everyone's Claude knows next — no re-explaining, no drift.
The CEO works out the new positioning at 1am, chatting with their Claude.
By morning, every teammate's Claude is already building from it.
The CTO asks Claude why you shouldn't use that library — the answer stays in their chat alone.
Now every developer's Claude already knows, and stops suggesting it.
A decision you talked through with Claude on Monday contradicts what someone ships on Thursday.
Now that decision follows the work — Claude.ai into Claude Code, automatically.
Three steps, about two minutes — then it runs on its own.
1Connect Relay to your team's Claude — no SDK, no manual config.
2What one teammate decides in Claude is saved for everyone, scoped by role.
3Your team's context follows them across Claude.ai and Claude Code — nothing gets lost.
We're a small startup, and we wanted plug-and-play shared memory for our team. The tools out there existed — but they were built for developers wiring memory into one agent, and they took real configuration to stand up.
The thing that pushed us: our CEO used Claude constantly and had sharp insights — but they lived in his Claude, and he never had time to write them up for the rest of us. Same with our PM. The best thinking on the team was trapped one person at a time.
So we built Relay for ourselves: shared memory native to the whole organization — not just dev teams, not just business — that you set up once and forget. Then we realized other teams had the same problem. So here it is.
— The Relay team
€99/mo
For your whole team — up to 20 members. No per-seat pricing.
Free for 7 days. Cancel anytime before it ends.
Let's talk
More than 20 members, or need something custom? We'll tailor a plan.
Relay is an MCP server your Claude already speaks — no SDK, no agents in your repo, no prompt engineering. Tenant-isolated, no code access, fully auditable.
{
"mcpServers": {
"relay": {
"type": "http",
"url": "https://sharedbrain.team/api/mcp",
"headers": { "Authorization": "Bearer <org-token>" }
}
}
}Claude's built-in memory is per-person and only works in Claude.ai. This adds the missing layer: decisions made in Claude.ai automatically appear in every developer's Claude Code, and a single admin can see and control everything the team's Claude knows.
| Claude's built-in memory | Shared team memory | |
|---|---|---|
| Who sees it? | Only you | Everyone on the team |
| Works in Claude.ai? | Yes | Yes |
| Works in Claude Code? | No | Yes |
| Admin visibility? | No | Full audit via Memory Browser |
The Memory Browser lets any organisation admin view, edit, or delete every memory your team has added. Nothing is hidden — you have full visibility and full control.

Yes. Each organisation's memory is fully isolated — no data ever crosses between teams.
Both are supported from the same memory pool. The Mac app wires both automatically.
Yes. Organisation admins can export memories from the Memory Browser in the dashboard.
Your subscription stops immediately. Memory data is retained for 30 days so you can export it. After 30 days, all data is permanently deleted.
Not yet. The service is MCP-compatible but currently optimised for Claude Code and Claude.ai. Support for other MCP hosts (Cursor, Windsurf) is on the roadmap.
Everything. €99/mo for your whole team — up to 20 members, unlimited memory entries, full access for both Claude.ai and Claude Code. Starts with a 7-day free trial; you can cancel anytime before it ends and won't be charged.
Stop starting from zero. Start your team today.
Start your 7-day free trial