Product Updates
July 2026 product update: Vantage reaches the dashboard, and Lab Reports returns layered results
July 2026: order blood and DNA panels from the dashboard with Vantage, and Lab Reports now returns layered results, report panels and coded interpretation flags.
TL;DR: Vantage arrived in the dashboard to improve observability and UX, so you can order blood and DNA panels and receive results as structured data without a lab relationship of your own. And Terra's agent skills and documentation MCP server let an AI coding agent build against accurate, current information.
Vantage: order blood and DNA panels from the dashboard
Vantage reached the dashboard in July with a full ordering flow so creating your own blood experience is not easier than ever before: a wizard for placing an order, a view of orders in progress, and the results when they arrive. It covers both at-home kits and in-lab collection, and the results come back in FHIR, the standard format clinical systems exchange data in.
The catalogue is discoverable, including which panels are available in which countries, and there is a test mode so you can build the whole flow, including the webhooks that fire as an order moves through fulfilment, before you place a real order. Orders are charged before shipment, and the order lifecycle is documented end to end so you know what each status means.
If you already have lab data in the form of PDFs, Lab Reports remains the way to structure it. Vantage is for when you want to originate the test. The documentation is at the Vantage API docs.
Give your AI coding agent accurate Terra knowledge
If you have tried to get Claude Code, Cursor, Codex or GitHub Copilot to write a Terra integration, you will have seen it produce something that looks right and is subtly wrong: an endpoint that used to exist, a field name from an older version, a webhook shape it half remembers. That is not the agent being careless, it is the agent working from training data with a cutoff.
July shipped two things that close that gap. The first is a set of Terra agent skills, which follow the open Agent Skills standard and work with any agent that supports it. Installing them is one command, npx skills add tryterra/agent-skills, which detects your installed tools and configures each of them. The skills teach the agent how to build with Terra correctly: not only how to set up the integration, but how to handle the data afterwards, which is where most of the mistakes actually are.
The second is a documentation MCP server. MCP, the Model Context Protocol, is the open standard that lets an AI tool query an external source while it works. Pointing your agent at Terra's documentation server means that when it needs an exact field name it looks it up rather than guessing. They work well separately and better together.
The dashboard gained a tabbed setup modal covering both, so getting your agent configured is a copy and paste.
Lab Reports now returns layered results and report panels
A lab report has more structure than a flat list of biomarkers, and our July's rework of Lab Reports preserves it: Results are now layered into the source line as printed, the biomarker it maps to, the measurement with its units, and the interpretation, which is the laboratory's own judgement of whether a value is normal, high or low. Where a report prints no interpretation, Terra computes one from the reference range and marks clearly which of the two you are looking at.
Panels are now modelled as entities in their own right, so a full blood count arrives as a panel containing its constituent measurements rather than as fifteen unrelated rows. The lab report webhook event is also now publicly documented. Everything is in the Lab Reports documentation.
Also in July
- A new
health_observation data type was added end to end, for observations that do not fit the existing daily, sleep, activity or body shapes. - The activity types reference was published, including newly distinguished types such as trail running and gravel cycling. It is at activity types.
- The dashboard gained an in-product status page, an AI-first support panel that can raise a ticket itself, and filtering and sorting on payload history including search by summary id.
- Fitbit: nutrition is delivered the moment a user logs food, and sessions recorded on a Pixel Watch are classified from the session's own sport type, so a Pixel-recorded run reaches you as a run.
