TEAM^AI · Workshop

Your team's best AI work lives in one person's session.Make it the team's. Build on it. Own it.

OpenX is where your team's workbenches connect - turn what each person does with AI into shared, reusable tooling the whole team builds from and owns. Connect your information, crystallize the work that repeats into shared sources of truth, and ship real apps. Every connection stays yours.

The problem

Your AI is sharp. It just doesn't know what you know.

The work that repeats - the good query, the right surface - lives in one person's session, not where the team's AI can reach it.

It can't see your data, your systems, your sources of truth - so it guesses, and you re-explain.

And wiring AI to real systems is the part nobody wants to own, or hand to a vendor.

OpenX makes that work the team's - and keeps it yours.

Own your connections

Connect your information to AI. Own every connection.

Wire your data, systems, and tools to AI - simulators, live devices, databases, external APIs. It all runs in your tenant, git-backed. Fire us, keep running: the connections, the queries, and the history stay yours.

What you build stays yours, not the vendor's.

Shared sources of truth

Crystallize the work that repeats.

Azi is your OpenX workshop assistant. Ask her in plain English - she configures flows, answers questions, and talks with you about your data. When a query is right, save it. It becomes a warm query - a named, team-reviewed source of truth that's automatically an API and an MCP tool. One person's good answer becomes the answer every teammate's AI builds from. No more five people, five versions of the truth.

Build real things

Turn it into APIs, interfaces, and apps.

Every crystallized query is an endpoint and an MCP tool - feed dashboards (Power BI, Grafana), build application interfaces, or wire it into Claude, Cursor, Copilot, or app-makers like Dify. Compose connections, surfaces, queries, and data flows on the canvas into real applications.

What's in the workshop

OpenX ships full ofworkbenches.

Every capability you build with, in the box. Bring your own, too.

Connect data

Stream from simulators, live devices, databases, and external APIs. No custom pipelines.

Surfaces

Shape your data so any AI can actually understand it.

Warm queries with Azi

Ask in plain English. Azi turns it into a query you can trust, and shows its work.

APIs

Save any query as an API. Power BI, Grafana, dashboards, workflows - instantly.

API Explorer

See and call your team's APIs directly, no guesswork.

Team

Bring teammates in. The right people see the right workbenches.

Security

Access rights per workbench: who sees what, who accepts what.

Workflow

Turn reviewed proposals into repeating, structured work.

MCP

Bring any AI to your workshop. Every workbench speaks MCP.

Move fast, safely

When it's consequential, AI proposes - your team decides.

Not everything needs a gate. But for changes that touch production, access, or shared truth, every decision is remembered, so nothing ships without your team's review.

01

Propose

AI drafts changes as structured proposals. It never acts alone.

02

Review

The right teammate reviews and accepts, per their access rights.

03

Ship

It ships because a human said so. Every decision is remembered, so your team's judgment compounds.

No meeting required

Review on your team's time.

A teammate crystallizes a new query - or a whole workbench - and proposes it. Someone reviews it the next morning, and the team adopts it. It becomes shared tooling everyone builds from. The workshop holds the record of what changed and why.

One answer becomes the team's

One person discovers it. The whole team owns it. It compounds.

When one person finds the query, the surface, the answer that works, the team reviews it and it becomes shared - a shared vault, shared sources of truth every teammate's AI draws from. Start solo and add teammates whenever; nothing to re-architect.

You built compounding context for yourself with fAI; now your team builds one together.

Capabilities compound

One person builds it.The whole team installs it.

When a workbench works, it doesn't stay in one person's setup. The team installs it with one command - not copy-paste, not a doc to follow. Capability compounds across the team.

Build your own

And you can build and host your own workbench.

Create a capability. Package it. Host the MCP server. Your team installs it with fai mcp @workbench-name and uses it from Claude, Cursor, Copilot - any AI.

The workbench we dogfood: the cognitive-loop skills workbench.

$fai mcp @cognitive-loop

Build your own workbench, host it as an MCP server, share it - one command and your whole team has it.

Connect what you know. Build what you need. Own all of it.

Start with fAI.Bring it into OpenX.

fAI is your workbench. OpenX is the workshop where your information, your team, and your apps come together - owned by you.

$deno run -A jsr:@fathym/fai/install
$fai mcp @workbench-name