Nexrall Canvas — Overview
Nexrall Canvas is the writing agent. Where Fullstack Agent Coding produces a deployed app and Nexrall Code edits your real repo, Canvas produces one self-contained document per session — a report, a slide deck, a chart pack, a dashboard, or long-form writing. One polished artifact, exportable to PDF.
Under the hood it's the same kind of agent: it reads, writes and surgically edits content with the same read-before-edit / unique-match discipline as the coding agent. The difference is that the "project" is a single HTML document rather than a file tree, and instead of bash and a test runner it has tools for charts, images, web search and export.
How it's different from Fullstack Agent Coding
| Fullstack Agent Coding | Nexrall Canvas | |
|---|---|---|
| Output | A deployed app with a backend and database | One self-contained document |
| Lifecycle | An ongoing project you keep extending | A finished artifact you export or share |
| Tools | File tree, shell, tests, deploy | Read/write/search the document, sources, charts, images, web search, export |
Canvas is deliberately simpler: one artifact per session, no multi-file project, no server. If a document needs many sections — a long report, a multi-slide deck — the agent builds that as one document with multiple pages, not multiple files.
What it can produce
Canvas writes real HTML/CSS/JS, not a fixed template, so the output is limited by what a web page can render — which is a lot:
- Documents — reports, contracts, letters, invoices, proposals, resumes, brochures, posters, catalogues, lesson plans.
- Slide decks — Reveal.js decks that read like a document in the app and become a true slideshow in fullscreen.
- Dashboards & interactive tools — live charts, KPI counters, filterable tables, calculators.
- Rich visuals — data charts, flowcharts, mindmaps, maps, 3D scenes, QR codes, barcodes, math notation, syntax-highlighted code, image before/after sliders.
See Charts, diagrams & 3D for the full library ecosystem.
What makes Canvas trustable
Canvas doesn't just claim a document looks right — every write is verified against a real render:
- Syntax + render checks catch broken JavaScript and blank pages on every write, not only when you happen to open the file.
- A real headless browser loads any CDN chart/3D library you used and confirms it actually drew pixels — a chart that stayed blank is flagged before the turn ends.
- Page-fidelity checks make sure a multi-page document exports with the page breaks you see on screen, not drifted or split mid-heading.
That machinery is covered in How it works and Security & trust.
Where to go next
- Building your first document — a walkthrough from prompt to artifact.
- Editing: by chat & in place — refine the document through chat or directly on the page.
- Sources & grounding — ground the agent in your own files, including scanned PDFs.
- Charts, diagrams & 3D — the visual libraries Canvas can reach for.
- Exporting to PDF — page sizes, orientation, and how export works.
- Sharing & managing artifacts — invite, public links, and your artifact library.
- How it works — the verification layers, model routing, and export engine under the hood.
- Security & trust — sandboxing and how Canvas treats your data as untrusted.