Veille connects to your existing ML platforms to discover and monitor AI systems without disrupting production. All connectors are read-only: no write access, no data extraction, no agent running inside your environment. Connector credentials are encrypted at rest and never logged.
Connector roadmap
| Integration | Status | What it reads | What it never reads |
|---|---|---|---|
| REST API | Available | Any system data you push programmatically | N/A, you control the payload |
| Manual intake form | Available | System metadata entered by your team | N/A, structured interview only |
| MLflow / Databricks | Beta | Registered model names, run metadata, production flags | Experiment data, feature tables, raw datasets |
| AWS SageMaker | Beta | Endpoint names, model metadata, tags, deployment status | Model weights, training data, inference logs |
| GitHub | Beta | Repository and Actions metadata for model artifacts and ML workflows | Source code, secrets, private repository contents |
| Azure Machine Learning | Q4 2026 | Workspace model registry, endpoint names, deployment config | Model weights, training data, inference results |
Integration roadmap is shaped by design partners. If your ML platform is not listed, contact us. The next connector is chosen based on what the founding cohort needs most.
Integration principles
Starting point
Start manual, then switch on the beta connectors when you're ready. The guided intake form covers everything the beta connectors discover, at the cost of a few hours of manual work upfront. Most organizations start manual and add connectors once the registry is established.
Book a call →