AGENTERACT
Let your agents interact...The SDK that connects AI agents to running apps.
Getting Started with Agenteract
For the complete installation guide, see the full documentation
Package manager:
Installation
First, install the Agenteract CLI. This tool manages communication between the AI agent and your local development servers.
npm install -g @agenteract/cli
Next, install the appropriate package for your project type:
For React Native (Expo):
npm install @agenteract/expo
For React (Vite):
npm install @agenteract/react
Demo Video
Check out this amazing demo that highlights the core of our developer experience — sleek, fast, and visually stunning.
Built on a comprehensive ecosystem of packages.
- @agenteract/coreCore protocol & utilities@agenteract/viteVite integration@agenteract/agentsAGENTS.md installer
- @agenteract/reactReact/RN bindings@agenteract/cliUnified CLI@agenteract/domWeb DOM utilities
- @agenteract/expoExpo bridge & CLI@agenteract/serverAgent runtimeagenteract-swiftiOS/Swift package
Agents That Actually Understand Your App
Most "AI agents" rely on vision or accessibility APIs to understand an app's UI. That approach is heavy, slow, and unreliable.
Agenteract flips the model — apps self-report their structure and controls in a simple, semantic format. Agents no longer guess what's on screen; they can query the actual component tree and call meaningful actions.
Example Flow
An example interaction flow where an agent queries the app’s UI hierarchy.
Agent requests
{ "action": "getViewHierarchy" }App responds
{ "status": "success", "hierarchy": { "root": "main(RootComponent)", "children": [...] } }Agent performs action
{ "action": "press", "target": "loginButton"}App executes bound handler and returns
{ "status": "ok", "result": "Button pressed" }Agent Registry and Commands
The agent maps component actions (for example, onPress or onChangeText) to callable handlers and commands in the form of JSON messages, such as:
{ "action": "tap", "target": "loginButton" }.Bridge
A lightweight WebSocket endpoint inside the app (AgentDebugBridge, AgenteractBridge, etc.) that sends UI snapshots and receives action commands.
<5KB
Bundle size
~0ms
Overhead
Multi-Platform
React Native, Expo, React web, Vite, Swift. One protocol, multiple platforms. Build agents that work across your entire stack.
Example Flow
An example interaction flow where an agent queries the app’s UI hierarchy.
Agent requests
{ "action": "getViewHierarchy" }App responds
{ "status": "success", "hierarchy": { "root": "main(RootComponent)", "children": [...] } }Agent performs action
{ "action": "press", "target": "loginButton"}App executes bound handler and returns
{ "status": "ok", "result": "Button pressed" }Agent Registry and Commands
The agent maps component actions (for example, onPress or onChangeText) to callable handlers and commands in the form of JSON messages, such as:
{ "action": "tap", "target": "loginButton" }.Bridge
A lightweight WebSocket endpoint inside the app (AgentDebugBridge, AgenteractBridge, etc.) that sends UI snapshots and receives action commands.
<5KB
Bundle size
~0ms
Overhead
Multi-Platform
React Native, Expo, React web, Vite, Swift. One protocol, multiple platforms. Build agents that work across your entire stack.
Build with your favorite frameworks
Your agents can interact with multiple app frameworks. More coming soon.
Contact Us
Have questions? We're happy to help!