Document Desk Agent
Document workflows, with people in control.
Document Desk Agent is an experimental WebMCP-powered tool for recurring document tasks. It explores how people and AI agents can work with documents together.
Independent development · Submitted September 2026 · Profile: UM-PT Mueller
Document workflow explored by the project
- Analyze PDF documents
- Identify document types
- Identify tasks and deadlines
- Suggest useful filenames
- Recommend filing locations
- Create short document summaries
- Highlight important information
- Let users review and approve actions before execution
The public demo uses a sample invoice and predefined analysis to demonstrate the approval workflow. It does not rename, move or modify files on disk. This is an experimental development project, not a finished document-processing product.
Human-in-the-loop
The AI agent prepares suggested actions. The user reviews them and decides which changes are applied. Proposals and execution remain separate.
Technologies
- WebMCP
- JavaScript
- HTML / CSS
- PDF.js
PDF.js is listed in the challenge submission. The currently published sample demo uses only HTML, CSS and JavaScript, with no external dependencies.
Challenge submission & project links
Submitted to the OpenAI WebMCP Challenge 2026 in September 2026 under the profile UM-PT Mueller. This is an independent project submission; no award, official partnership or certification is claimed.