Ask questions about a project — and let the agent do the mapping, with your approval.
The AI Agent sits beside whatever you are working on. Open it with the sparkle button in the top-right corner; it picks up the project you are in, so questions do not need much setup.It can read your project data and, when you approve it, change your project.
The agent is connected to this documentation, so “how do I…” questions get answered from the guide rather than from guesswork.
A question about the product
You can see it work: the trace shows a search_real_time_lca_docs step, and the answer arrives with the steps to follow.
An answer, with its sources
Underneath sits a Sources list linking the pages it drew on, so you can jump to the full article and check the answer against it. That makes it a reasonable first stop for questions you would otherwise search the docs for — and it is offered as exactly that from the Feedback menu.
Docs answers are generated too, and the same caution applies: follow the sources when the answer matters. Where the documentation is out of date, the agent will be as well.
Skills are reusable instructions the agent can invoke. Attach one from Skills in the composer and it shows as a chip on your message — the chat above uses rtlca-map-material.
The skills library
Curated skills are maintained by Real-Time LCA and cannot be edited, but you can duplicate one and adapt it. New skill writes your own from scratch.
The agent shows its work. For a mapping request it resolves the materials, reports what is blocking them, and finds candidate EPDs before proposing anything.
What it found
Notice it flags what is missing — here, two materials with no lifetime set — rather than guessing.
The agent cannot write to your project unsupervised. Anything that changes data pauses for Approval needed, naming the exact operation.
Approval needed
Allow lets that operation through; Deny stops it. Above the prompt you can see every tool it has called.
A recommendation, awaiting confirmation
It states what it is about to do — which EPD, for which materials, and why — and waits for a yes.
Done
Afterwards it summarises what changed: the EPD used, the lifetime it set, how units were converted, and how much mapping coverage went up. Every one of those changes is a normal entry in History and can be restored.
The AI Agent section in the left rail has two more pages.
Usage
Usage tracks tokens and sessions across your workspace by month, broken down by feature.
History
History lists every session across the workspace — who ran it, which feature, when, and how much it consumed. Both can be exported.
Responses are generated by AI and may be inaccurate. The agent can read your project data and make changes to it — review anything it proposes before relying on the numbers. See Transparency and privacy.
