Chat that knowsthe atlas.
The Agent is a render-aware chat interface that reads the current atlas state and runs presets on your behalf. Ask 'show this in winter at dusk' and the Agent picks the right preset chain, queues the runs, and drops the outputs as nodes in the atlas — with the same credit accounting and lineage as a manual run. Three model tiers (fast, normal, high) and per-message size and quantity controls keep cost and fidelity dial-able.
How it will work
- 01step 01 / 03
Open the chat next to the atlas
The Agent panel docks alongside the canvas. It reads the atlas state — which renders are present, which presets have run, which nodes are selected — so prompts can reference 'this view' without re-uploading.
- 02step 02 / 03
Ask in plain English
'Relight this for golden hour and add three figures on the terrace' — the Agent translates that to L.02 + E.01 with the right parameters, queues both runs, and reports cost before firing.
- 03step 03 / 03
Outputs land in the atlas
Every response materialises as nodes in the atlas, with edges back to the source. Inspect the prompt the Agent built, fork it manually, or undo the run — the chat keeps the receipts.
Presets in this mode
Each preset is named, parameterised, and routed to the engine that fits the task.
What it's for
Voice-of-the-brief variations
Architect dictates the brief: 'autumn morning, modest fog, two pedestrians'. The Agent generates three variants matching the description without the architect touching a slider.
Onboarding without the manual
New hires don't need to learn 30+ preset codes. Type the intent in English, the Agent picks the right preset and explains why. The studio's atlas knowledge transfers in days, not months.
Bulk operations from one prompt
'Apply L.01 Day → Dusk to every source render in this project.' The Agent fans out the runs in parallel, reports total credit cost upfront, and drops the variants as a row of nodes.
Conversational cost-and-tier control
Switch the Agent between fast (1 credit, 3 s), normal (2 credits, 8 s), and high (3-5 credits, 15 s) per message. Discuss tradeoffs with the chat itself.
Frequently asked
01What is the Agent in Renovato?
What is the Agent in Renovato?
The Agent is Renovato's render-aware chat interface. It reads the current atlas state, translates plain-English requests into preset chains, queues the runs, and drops the outputs as nodes — so an architect can drive the platform without touching the canvas directly.
02Which models power the Agent?
Which models power the Agent?
The Agent runs on tiered LLMs — Anthropic Claude, OpenAI GPT, and Google Gemini — selected per message via the fast / normal / high tier control. The image and video generations the Agent fires are still routed across Renovato's full vendor stack (OpenAI, Gemini, ByteDance, Kling, open source).
03How is the Agent different from ChatGPT or Claude.ai?
How is the Agent different from ChatGPT or Claude.ai?
The Agent reads your atlas — which renders are loaded, which presets have run, which nodes are selected — so prompts can reference 'this view' or 'the dusk variant' without describing them. It also writes back into the atlas as new nodes with full lineage; ChatGPT and Claude.ai are general chat surfaces that can't see or modify your project state.
04Does the Agent run presets without asking?
Does the Agent run presets without asking?
No. The Agent reports the planned preset chain plus total credit cost before firing. You confirm, then it runs. For long bulk operations the Agent shows progress and lets you abort mid-chain.
05What are the per-message tiers?
What are the per-message tiers?
Fast tier (Q.01) uses smaller LLMs for quick replies — 1 credit per message. Normal tier (Q.02) is the default — 2 credits, better at complex preset chains. High tier (Q.03) uses frontier models for nuanced creative direction — 3-5 credits per message.
06Can the Agent explain a preset's parameters?
Can the Agent explain a preset's parameters?
Yes. Ask 'what does L.02 Golden hour change?' and the Agent walks through the parameters, the model it routes to, the typical credit cost, and a side-by-side example output.
07Is the Agent included in every plan?
Is the Agent included in every plan?
Yes. Agent access is included on every Renovato tier from Starter ($19/mo) up. Per-message credit costs come out of the same credit pool the rest of the platform uses.
