iPolloWork: a local-first visual AI workbench for creating editable code, documents, and multimedia content via agents

What it solves

iPolloWork provides a local-first visual workbench for AI agents, moving beyond simple chat interfaces. It solves the problem of "final-file" outputs by allowing agents to generate editable code, documents, presentations, websites, designs, and videos within a single workspace, rather than just providing a static file or a chat transcript.

How it works

The system operates as a desktop application (built with Electron) that integrates a local server and a runtime orchestrator. It uses a "sidecar" architecture where it connects to OpenCode for execution. Users describe a goal, and the agent plans and executes the task using tools, local files, and browser automation. The workspace is local-first, meaning users can bring their own models or providers and maintain control over permissions and execution on their own machine.

Who it’s for

It is designed for developers and creators who want a Codex-like agent experience but with a visual, editable workspace for a variety of media types (not just code), and who prefer local control over their data and execution environment.

Highlights

  • Agent-first execution: Agents can plan work, run commands, and modify files based on the current state.
  • Editable results: Outputs like websites, decks, and videos remain editable after generation.
  • Local-first control: Runs on the user's machine with support for custom models, MCP servers, and plugins.
  • Multi-format support: Handles repositories, local files, browser tasks, documents, and visual designs in one place.

Sources