Platform
Run ATE·IQ on your own workstation, point it at the LLM provider your organisation already contracts, and decide what else it connects to. Specs, programs, boards and production data are stored in a local database, and the app's services listen only on the machine itself. Your project data leaves only where you point it: prompts to your own LLM provider, on your own keys, and any git repository, Jama Connect server or external device-model endpoint you connect.
ATE·IQ is a desktop product built from web technology, not a cloud service with a desktop wrapper. The IDE window and the IG-XL agent beside Excel are served by the same local server and work the same projects.
The desktop shell lives in the system tray. Closing the main window hides it there instead of shutting the product down, so the IG-XL agent keeps working beside Excel. A workbook ingested from Excel becomes part of the project the IDE sees, with no export step.
ATE·IQ ships no model and holds no keys of its own. Configure the provider you already use, supply your own credentials, and every AI feature routes through that choice, so your existing egress policy and data agreement with that provider apply unchanged.
| Provider | Configuration |
|---|---|
| Anthropic | Direct API with the customer's key. |
| OpenAI-compatible | Any endpoint speaking the OpenAI API — hosted, or a self-hosted gateway inside the customer's network, via a configurable base URL. |
| Azure OpenAI | Endpoint, deployment and API version — for organisations whose model access is already contracted through Azure. |
Switch provider at runtime from Settings, with no reinstall. See which project artefacts a question will draw on before it goes: the Context Curator previews the assembled project context, artefact by artefact with its score, and lets you choose what is included. The validators, analyzers, DIB checker, STDF tools and traceability reconciles are deterministic and run with no provider configured at all.
Connect the systems your team already runs. Every connection is opt-in, and each one is the only traffic it adds.
| Integration | What you get | Where traffic goes |
|---|---|---|
| Git repositories | Clone and sync test-program and board repositories into a project; the programs, board exports and STDF files in them are indexed and imported. | To the repository host you connect. Access tokens are stored encrypted. |
| Jama Connect | Pull requirements from a Jama Connect project into a structured spec. | To the Jama Connect server you configure, over HTTPS. |
| External device model | Judge generated programs against a device model you run on your side, over ATE·IQ's own external-model protocol. Your model stays with you. | To the endpoint you register on the project. Its secret is stored encrypted and never shown back. |
| IG-XL knowledge in your AI assistant | A local Model Context Protocol (MCP) server: an MCP-capable assistant on the engineer's machine, such as Claude Desktop or Claude Code, calls the same IG-XL knowledge tools the IG-XL agent uses. | None added by ATE·IQ: the server talks to the assistant over standard input and output and opens no network port. |
The MCP server takes one command: npm run igxl:mcp starts it, and a
single claude mcp add line registers it with Claude Code, which then
launches it whenever it needs the tools. It is an easy first step for a team that
already works in an AI assistant: built-in IG-XL and UltraFLEX platform
knowledge, from the tool they already have open.
| Control | Implementation |
|---|---|
| Network exposure | Local-only by design — services are reachable only from the machine itself, and nothing is exposed to your network beyond what you configure. ATE·IQ has no cloud component of its own. |
| Stored secrets | Encrypted at rest with AES-256-GCM: an LLM API key entered in Settings, Git access tokens, Jama Connect tokens, a device-model endpoint's secret and the password for the IG-XL simulator VM. None of these is stored in plain text; OpenAI-compatible and Azure credentials are read from environment variables on the machine. |
| Git tokens on disk | A token is handed to git for each command and never written into a cloned repository's configuration. |
| Sign-in | Repeated failed sign-ins are throttled. |
| Excel link | The link between Excel and ATE·IQ runs over localhost, and every call carries a per-user secret generated on first run — a request without it is refused. |
| Workbook file reads | Workbook access is constrained to the user profile and OneDrive roots and to .xlsm/.xlam extensions — it cannot be used to read arbitrary files. |
| API authentication | Every page and every API surface authenticates the caller in its own right, and the product's internal calls are authenticated as well — there is no unauthenticated back door between components. |
Test-program generation, code-map indexing, and document and board ingest run in a background worker without blocking the app. A failed job is retried, a job that still fails is shown with its error rather than dropped, and work interrupted by a crash is picked up again on restart. The status bar keeps checking the platform's health while you work, so a stalled pipeline is a fault you can see.
| Part | Today |
|---|---|
| Excel add-in | Installed per user with no administrator rights: one add-in file in the user's profile and two per-user registry settings. An ATE·IQ tab in Excel's ribbon opens the IG-XL agent beside the open workbook; in plain Excel nothing is configured per test program. |
| Desktop app | Developer-grade to install: it runs from a source checkout and expects Node.js and a local PostgreSQL on the machine, both bound to the machine itself. |
| Packaged installer | Not shipped yet. The direction is a single packaged Windows executable with an embedded data store, removing those prerequisites — a direction of travel, not a dated commitment. |
| LLM provider | Any of the three above: choose the provider in Settings, where an Anthropic key is entered; an OpenAI-compatible or Azure endpoint is set by environment variables on the machine running ATE·IQ. A self-hosted OpenAI-compatible endpoint keeps prompts inside your network. |
| Excel on another machine | Excel in a virtual machine or on a second PC connects with a copy of the per-user bridge secret and a network path to the app; nothing binds to the network by default. |
Screenshots in this guide were produced on a single Windows workstation from its local database: the demonstration project (a spec, the IG-XL program generated from it, and nine STDF files of seeded synthetic lots with planted defects) and, separately, a real Allegro 17 board read from its Fabmaster export.