Canvas is where your team and the agents reason together. Not a chat. Not a copilot. An investigation surface — where every question ends in an artifact: a brief, an RCA, a post-mortem.
∞
3 → 1
100%
Canvas reads from a live graph of your environment — services, dependencies, owners, SLOs, recent deploys, on-call rotations, runbooks. Not a snapshot. Not a CMDB you forgot to update. A graph that updates itself from your live runtime, your IaC, your incident history.
When you ask a question, the agent traverses the graph. The same graph Sentinel uses to score signals. The same graph Resolve uses to choose runbooks. One source of truth.
Every Canvas investigation ends in an artifact. Not a chat transcript. Not a summary. A finished document — RCA, brief, post-mortem — formatted, sourced, ready to share.
Your team approves it. Your auditor reads it. Your post-mortem template doesn't change. The agent did the writing. You did the deciding.
Canvas treats every investigation as a tree. Don't agree with the agent's conclusion? Branch from the step you disagree with, redirect the reasoning, run it again. Compare branches side-by-side.
Your most thoughtful investigators stop reviewing summaries and start guiding chains of thought. The agent learns the shape of how your team actually reasons.