Enterprise Claw Agent

Tasks

• Validate Workstation — check dependencies, create Python environment, install libraries, extract agent runtime.
• Run Agent — execute an autonomous AI agent with a plain-English goal, configurable model, and optional tool toggles.
• Browser & Desktop Automation — navigate web apps via Playwright, fill forms, click buttons, or control desktop apps via
• Persistent Memory & Custom Tool Library — remember facts across runs, create reusable tools on the fly, and build an eve
• Document Integration — read PDFs (text + scanned via OCR), spreadsheets, and text files to extract and process data.

**Special Access Preview! This AI agent package from Automation Anywhere Labs is a forward-leaning preview of what we’re working on.

Enterprise Claw Agent brings autonomous AI agent capabilities to Automation Anywhere. Give it a goal in plain English, and it figures out the steps — reading documents, browsing the web, writing files, executing code, and filling out forms — all without manual flow design.

The Problem: Building automations for dynamic, multi-step business tasks is time-consuming. Each workflow requires custom logic, error handling, and maintenance. When processes change, bots break and need rework.

The Solution: Enterprise Claw Agent deploys an intelligent agent directly into your AA workflows. The operator provides a goal (e.g., "Read the receipt PDFs in the receipts folder and create an expense report spreadsheet"), selects a model, and the agent autonomously determines how to accomplish it — calling built-in tools for PDF reading, file writing, shell commands, Python execution, and more.

The agent produces a dictionary output containing the status, final response text, run ID, log path, and any custom tools it created. Detailed audit logs (human-readable and structured JSONL) capture every reasoning step, tool call, and token cost for full observability.

Use cases include expense report processing, document generation, data entry, research and analysis, file batch processing, and system validation. The included demo bot with three sample receipts lets you see it in action immediately. Both Windows and macOS workstations are supported with a cross-platform supported package.

Full project source code: Automation Anywhere GitHub