Developer releases Panel, an open-source research workspace with agent-built UI
The new local tool integrates Jupyter notebooks and PDFs into a dockable interface where AI agents can programmatically generate their own visualization panes.
Developer greentfrapp has released Panel, an open-source local research workspace designed to eliminate the constant context-switching inherent in technical analysis. The tool unifies disparate research elements—including AI chat, document readers, and live code—into a single, cohesive environment.
Panel is architected as a local web application utilizing a Vue and dockview frontend paired with a Python FastAPI sidecar. The interface supports dockable panes for markdown files, PDFs, and Jupyter notebooks with full kernel support. A core technical feature is the "Module" system, which employs typed inputs, outputs, and intermediates. This structure allows for complex inter-module workflows and provides observability for researchers monitoring long-running computational jobs.
The friction of research
Modern research typically requires users to juggle a fragmented stack of tools. A researcher might simultaneously maintain an LLM chat for brainstorming, a PDF reader for literature review, a code editor for implementation, and various external visualizers for data analysis. According to greentfrapp, Panel was created specifically because research work requires constant switching between these different modalities, which often disrupts the cognitive flow of the analyst.
A dynamic agentic interface
What distinguishes Panel from a standard IDE is the role of its integrated AI agent. Rather than acting as a simple chatbot, the agent can programmatically modify the workspace UI to better present findings. On request, the agent can create custom visualization panes and apps to handle data that built-in viewers cannot. Examples of these agent-generated components include specialized SQLite browsers and protein structure viewers.
By allowing an agent to generate its own UI components on the fly, Panel moves toward a more dynamic "agentic IDE." This shift means the interface adapts to the specific data being analyzed, reducing the need for users to export data to external third-party tools for visualization. This capability transforms the workspace from a static container into an active participant in the research process.
Deployment and availability
Panel is released under the MIT license. To install the workspace, users require Node 22.18+ (or 24.12+), pnpm, and the uv Python package installer. The project is currently available via GitHub for local deployment, allowing researchers to maintain full control over their data and environment while leveraging agentic automation.