Team Cloud / Founding pilot

Your agents should remember the same decisions.

Engraphis Team gives a small engineering team one shared, inspectable memory for coding agents. Approved decisions, conventions, incidents, and architecture rationale stay in one governed place instead of being rediscovered in five separate chats.

Apply for the founding pilot Compare plans

This is an application, not a checkout. We confirm fit and the exact pilot scope before any payment. Standard hosted plans include a 7-day Pro trial and a 14-day Team trial with 5 trial seats, both no-card and starting after email verification. The founding pilot is a separate written-scope engagement. The free Apache-2.0 local core remains independent.

The gap

Local memory is personal. Team knowledge is not.

A developer can install the free core and give one agent a durable memory. A team still loses the decisions that cross people, repositories, and agent sessions.

The repeated question

“Why did we choose this?” The answer lives in one engineer’s chat history, a stale pull request, or a meeting note. The next agent starts from zero.

The silent divergence

Two agents, two conventions. One follows the current deploy rule; another follows the superseded one. Nobody can see where the split happened.

The governance gap

Shared context needs a boundary. Teams need named seats, roles, an approval path, and an audit history—not an ungoverned shared folder.

The founding pilot

Five seats. One shared workspace. Direct onboarding.

We are onboarding a small number of engineering teams that already use MCP-capable coding agents and have real project context worth preserving.

Founding Team

Team Cloud pilot

$1,000 / year

Five named seats at the annual Team rate of $200 per seat. Review the annual Team checkout after the written pilot scope is accepted.

For 3–5 person product teams using Claude Code, Codex, Cursor, Cline, Windsurf, Zed, or another MCP client.

Five named seats with owner, admin, member, and viewer roles
One shared workspace with an explicit server-readable hosting choice
Review and approval before a memory is eligible for shared recall
Organization audit history and scoped agent credentials
Optional encrypted Cloud Sync, separate from hosting
60-minute onboarding and direct maintainer support during the pilot

Apply for the pilot

Scope

What we confirm first

Scoped

No payment until both sides accept the written scope

Your team and the context you want to share
The exact runtime, data region, and retention terms
Which workspace is server-readable and who approves memories
Support owner, response expectation, and exit path
Whether shared hosted recall is enabled for your pilot

Ask a question

Community

Free local core

Apache-2.0, no account required

The complete local recall engine
MCP server, dashboard, CLI, and library
Local history, graph, and manual consolidation
No account requirement or 24-hour write limit

Inspect the core

Availability. The Team Cloud pilot is an early-access offer. Shared hosted Team memory and hosted MCP are not generally available and are enabled only after the reviewed runtime passes the pilot’s acceptance gates. If we cannot enable the scope you need, we will say so before invoicing and will not take payment for an unavailable capability.

How it works

A short, attended path to one useful shared recall.

01 · Apply

Send the team size, the agents you use, and one piece of context your team keeps repeating. A short note is enough.

02 · Fit call

We confirm the workflow, the data boundary, and whether the current pilot can serve you. No generic demo.

03 · Onboard

The owner connects the workspace, approves one useful memory, and invites one teammate. We stay on the call until the teammate recalls it in a fresh agent session.

04 · Decide

Review the audit trail and the result. If the pilot is useful and the scope is accepted, the annual Team subscription begins through the authenticated annual Team checkout.

The return

One avoided rediscovery can cover a month.

This is an illustrative calculation, not a guarantee. Use your own loaded rate and time.

Example

Five engineers each lose 30 minutes a week to finding or re-explaining context: 10 hours a month. At an illustrative $75/hour, that is $750 a month of avoidable work. The annual pilot is $1,000.

What we measure

Time to one approved shared memory, time to a teammate’s successful recall, and the number of repeated explanations the team can point to. We do not claim a benchmark or productivity guarantee.

Boundaries

Local by default. Hosting is a separate choice.

The free core

The local engine, dashboard, MCP server, CLI, and library remain Apache-2.0 and usable without an account. The pilot does not take that away.

Hosted content

An owner explicitly chooses a workspace for server-readable hosting. Selecting hosting does not enable encrypted sync or managed processing; those are separate choices.

Review

New shared memories can require an eligible reviewer’s approval before agent recall. The approval is a browser action with a reason, not an MCP tool.

Exit

Authorized members can export retained workspace data. Removed members and suspended organizations remain denied. We agree the exit path in writing before the pilot starts.

Apply

Bring one real team problem.

Tell us the context your team keeps rediscovering. We will reply with fit, availability, and the next step. Do not include secrets, customer data, or credentials in the application.