v2.1.0 — AI Computer Use Inside
RocketMouse AI
The reliability of traditional RPA meets the intelligence of AI.
RPA just got a brain and eyes.
0
Blocks
0
Features
0
Unit Tests
0
AI Providers
The reliability of traditional RPA meets the intelligence of AI.
RPA just got a brain and eyes.
Powerful automation without writing a single line of code
Drag & drop blocks to connect them. Build macros intuitively with no coding required.
AI that sees, understands, and operates your screen. Autonomous task completion via Claude Computer Use API.
Control Edge/Chrome via Playwright. Automate web forms, data collection, and more.
18 Excel operations via COM. Read/write cells, set formulas, run macros, and manage sheets.
OpenCV + Windows OCR to detect images and text on screen, find positions, and interact.
Pause, step through, watch variables, and set breakpoints for reliable debugging.
App Screenshots — Coming Soon
Comprehensive blocks for every automation scenario
Click, drag, scroll, and more 16 blocks
Key input, text paste 2 blocks
Window management 8 blocks
Playwright browser automation 16 blocks
COM-based Excel operations 18 blocks
File and folder operations 15 blocks
Conditions, loops, functions 17 blocks
LLM API integration + Vision 1 block
Image recognition, text reading 9 blocks
JSON, regex, dates, lists 22 blocks
Define, evaluate, reference 5 blocks
App launch, ZIP, Base64, Hash 18+ blocks
RocketMouse AI is the only RPA tool that combines the best of both worlds
| Traditional RPA | AI Computer Use | RocketMouse AI Best of Both Worlds | |
|---|---|---|---|
| Execution Speed | ⚡ Fast (instant) | 🐢 Slow (LLM inference per step) | ⚡ Routine tasks run fast + AI only when needed |
| API / Running Cost | 💲 Zero (fully offline) | 💸 High (screenshot per step) | 💲 Routine tasks are free + minimal AI usage |
| UI Layout Resilience | ❌ Fragile (coordinate/selector dependent) | ✅ Strong (visually identifies elements) | ✅ AI visually recognizes UI elements |
| Unexpected Dialog Handling | ❌ Can only stop | ✅ Self-decides and adapts | ✅ AI Autopilot handles autonomously |
| Ease of Creation | ⚠ Manual coordinate picking | ✅ Natural language instructions | ✅ Drag & drop + natural language |
| Reproducibility | ✅ Same input → same result | ❌ Different path each run | ✅ Deterministic for routine + AI for judgment |
| Self-Healing | ❌ Not possible | ✅ Observes and self-corrects | ✅ Built-in self-healing |
| Offline Operation | ✅ Fully supported | ❌ Requires internet | ✅ RPA runs offline + local LLM support |
| Browser Automation | ⚠ Varies by tool | ⚠ Screen-only (no DOM) | ✅ Playwright (DOM) + AI Vision |
| Excel Operations | ⚠ Varies by tool | ❌ Screen-only | ✅ COM API (18 ops) + screen control |
Dual-provider Computer Use: Anthropic Claude and OpenAI GPT-5.5. RPA now has a brain and eyes.
"Click the Save button" — AI sees the screen, locates the element, and clicks it. No coordinates needed. 2-pass refinement detects even small icons at screen edges with high accuracy.
AI Vision"Open Notepad, type Hello World, and save" — Multi-turn autonomous operation via Claude or OpenAI GPT-5.5 Computer Use API. Takes screenshots, decides the next action, and continues until the task is complete. Switch providers freely.
Computer Use API (Claude / GPT-5.5)"Read the error message on screen" — AI Vision reads any text on screen and stores it in a variable. Unlike traditional OCR, it understands context and semantics.
AI Vision"Wait until the download dialog appears" — AI periodically checks the screen and waits intelligently until the condition is met. No more fixed-time waits.
AI Vision"Verify the file was saved correctly" — AI examines the screen after an action and returns true/false for conditional branching in your macro.
AI Vision"Is the login screen showing?" — Use AI judgment in if/while conditions. Build dynamic branching based on visual screen state, all with no code.
AI Vision5 LLM providers supported
Type "Open Notepad, type Hello World, and save" to automatically generate and place the corresponding block sequence.
OpenAI (GPT-5.5 series), Anthropic (Claude 4.6/4.7), Google Gemini, Groq (Llama 4 Scout, etc.), and local LLMs (LM Studio/Ollama). Choose your preferred provider.