rive-cli / Rive authoring for coding agents

Create interactive Rive animations. With the coding agent you already use.

Describe the experience you want. Your agent authors the scene, compiles a real .riv file, renders the result, and refines it — all inside your repo.

Code-first source · real .riv output · Rive runtime checks

MADE WITH RIVE-CLI 01 / 03
REAL
.RIV
Authored as JSON · compiled by rive-cli · playing live coffee_loader.riv

Why rive-cli

Rive is built for interaction, not just playback.

rive-cli gives your coding agent a code-first way to create the graphics, motion, and state logic together. The source stays readable in Git; the output is a real .riv file that can be loaded, rendered, and improved without leaving the product workflow.

An agent-native authoring loop

From an idea to an interaction you can ship.

The agent writes the source. The CLI turns it into Rive, shows the result, and gives the agent enough evidence to make the next pass better.

01

Describe

Ask for the experience in plain language. Your agent writes inspectable scene JSON.

scene.json
02

Build

Compile graphics, animation, and interaction into a real Rive binary.

rive-cli generate
03

Refine

Render key frames and inputs, inspect what happened, and iterate.

rive-cli render

Bring your own coding agent

Keep the animation work inside your existing workflow.

A coding agent that can edit files and run commands can use rive-cli. Templates provide a known-good start. The schema explains what is valid. Structured errors and text render summaries tell the agent what to change next.

Read the agent skill and schema ↗
AUTHOR.RUN / COFFEE_LOADERRIVE LOADS
$ rive-cli generate \
  parity/reproductions/coffee_loader.json \
  -o coffee_loader.riv

$ rive-cli render coffee_loader.riv \
  --frames 0,15,30,45 \
  --preview
RUNTIME RESULT 4 / 4 frames rendered · source stays editable

Quality stays in the loop

Create first. Prove it before you ship.

The verification lab loads files made by rive-cli in the Rive runtime and compares them with known references. That gives an agent more than a successful command: it can see whether the file loads and how closely it matches the intended result.

Open Verification Lab

Start in your own repo

Give your agent a Rive workflow it can run.

# install from this repo
$ cargo install --path .

# scaffold, build, and render a scene
$ rive-cli new spinner > scene.json
$ rive-cli generate scene.json -o scene.riv
$ rive-cli validate scene.riv
$ rive-cli render scene.riv --preview

Ask your agent to start from a template, inspect the schema, and edit scene.json. The source stays reviewable; the output is a real .riv file ready for Rive runtimes.

See every command and the agent skill ↗