# A dashboard in the browser, an agent on the bench.

> AutoST has three parts. The web dashboard is where you define ECUs (components), configure scans and read results. The backend orchestrates jobs, stores findings and issues API tokens. Carbyne is a bench agent that runs on Windows or Linux, connects to the ECU over CAN, CAN FD, DoIP or SOME/IP, and executes the scans, reporting progress and results back over HTTPS. Several agents can test several ECUs in parallel.

AutoST has three parts: a web dashboard, a backend that orchestrates and stores, and the Carbyne agent on your bench. See the scan flow from wiring an ECU to the fix.

Source: https://auto-st.com/how-it-works

AutoST splits into three parts: the web dashboard you work in, a backend that orchestrates scans and stores results, and Carbyne, the agent that sits next to your ECU and talks to the bus.

## What the bench looks like

One ECU, one interface, one tester running Carbyne, and the AutoST backend. Nothing exotic.

ECU (The target on your bench) --CAN / DoIP/HTTPS--> Vector interface (CAN, CAN FD or DoIP) --CAN / DoIP/HTTPS--> Carbyne agent (On a Windows or Linux tester) --CAN / DoIP/HTTPS--> AutoST backend (Dashboard in the browser)

The Carbyne agent polls the backend over HTTPS, so the bench needs no inbound ports. Swap the Vector interface for SocketCAN or a Panda, and run several benches in parallel.

## Dashboard, backend, bench agent

### Web dashboard

Define ECUs, configure and launch scans, watch progress live and read findings. Role-based access, groups and multi-tenant.

### Backend

Orchestrates the job queue, stores scans and findings, scores risk and issues the API tokens that drive CI/CD.

### Carbyne agent

Runs on your bench on Windows or Linux, connects to the ECU over the bus or IP, executes scans and streams results back over HTTPS.

## From a wired ECU to a fix plan

1. **Define the component** Create the ECU in the dashboard: its interfaces (CAN bus, Ethernet) and protocols (CAN, CAN FD, UDS, DoIP). A 3-step wizard walks you through it.
2. **Pair the agent** Install Carbyne on the bench machine and pair it with an API token. It reports an agent ID and shows up as an available target.
3. **Configure the scan** Pick the engine and mode, the adapter and bitrate, keep-alive, and any ODX/CDD or DBC files. The wizard previews the whole configuration.
4. **Run and watch** The agent executes the scan and streams progress: iterations, findings, requests per second. Several agents can run in parallel.
5. **Review and export** Read the findings, accept or comment on risks, and export a PDF report plus the fix plan as SARIF, CSV or JSON.

## It fits how you already work

AutoST does not need a lab full of security tools. One container runs the dashboard and backend; the Carbyne agent runs next to the ECU on hardware you already have.

Comments and risk acceptance carry over when you re-run a scan, so a fix shows up as resolved and the noise does not come back. Scan history per component gives you a record over the life of the ECU.

**Good to know**

- Agent on Windows and Linux
- Jobs over HTTPS, no inbound ports on the bench
- Parallel agents for parallel benches
- Risk acceptance and comments carried across re-scans

## FAQ

**Where does the testing actually happen?**

On the Carbyne agent, next to the ECU. The backend orchestrates and stores; the agent does the bus and IP communication. Nothing is tested from the backend directly.

**Can several ECUs be tested at once?**

Yes. Each bench runs its own agent, and the backend schedules scans across all of them in parallel.

**Does the bench need to be reachable from outside?**

No. The agent polls the backend over HTTPS and pulls its jobs, so no inbound ports are opened on the bench.

---
AutoST by Zyberum. Canonical page: https://auto-st.com/how-it-works
