Connect Virex Memory over MCP
By the end of this your coding agent will be writing what it learns into your organization memory and retrieving it in later sessions. Most of the work is account setup, and the connection itself is one configuration step.
What you need first.
Three things, and the third is the one people miss.
- An MCP-compatible coding agent installed and working, such as Claude Code, Cursor, Codex, Gemini CLI, opencode, or Cowork.
- A Virex account. If your team already has an organization, ask an admin to invite you rather than creating a second one, since each user belongs to exactly one.
- A Virex Memory seat assigned to you. Without it the tools will not resolve, and this is the most common reason a correct configuration appears to do nothing.
Five steps
Each step has a check, so you find out where it went wrong at that step rather than at the end.
- 01
Create your account and organization
One login covers every Virex surface. If your team already has an organization, join that one instead of creating your own, because memory is scoped to the organization and a second one would be a separate corpus.
Check: You can sign in and see your organization name in the portal.
- 02
Subscribe to Virex Memory
Checkout starts your organization subscription, or adds Memory as another line item if you already run Virex Remote. Either way your organization ends up with one subscription.
Check: Billing shows a subscription with a Virex Memory line item.
- 03
Assign yourself a seat
An admin assigns seats per member per product. One Memory seat covers every MCP client you use, so there is nothing to repeat per tool.
Check: The Memory product is no longer locked for you in the product portal.
- 04
Connect your MCP client
Point your coding agent at the Virex Memory MCP server. The exact configuration differs per client, and the docs carry the current form for each one rather than this page going stale.
Check: Your client lists the Virex Memory tools among its available tools.
- 05
Run a session and verify
Do a real piece of work rather than a test prompt. Agents write memories when they learn something worth keeping, so a trivial session may legitimately produce nothing.
Check: New memories appear in the memory browser, attributed to the project and the subagent that wrote them.
Common failure modes.
In rough order of how often they are the actual cause.
Questions about this guide
Which clients can I connect this way?
Any MCP-compatible client, including Claude Code, Cursor, Codex, Gemini CLI, opencode, and Cowork. The account steps are identical for all of them and only the client-side configuration differs.
Do I need a separate seat for each client?
No. Seats are per member per product. One Virex Memory seat covers every MCP client that member uses.
How do I know memories are actually being written?
Open the memory browser in the product portal. Memories appear there as your agents write them, attributed by project and subagent.
Nothing appears after a session. What went wrong?
Usually one of three things: the client is not actually connected, the member has no Memory seat, or the session did not produce anything the agent judged worth keeping. The troubleshooting section covers each.
Can I connect more than one client to the same memory?
Yes, and that is the point. All of them resolve to the same organization-scoped corpus, so a memory written from one is retrievable from another.
Give the next session a head start.
Subscribe to Virex Memory, assign a seat, and connect the client your team already uses.