Observatory

Inspiration

We built Observatory to make AI-powered data analysis more transparent and actionable. We started with a simple idea: instead of having an AI assistant only respond with text, we wanted the AI agent to actually investigate data and take actions inside the workspace.

We designed the application around a CSV dataset workflow. Users can upload their data, ask an investigation question, and let the agent explore the dataset, analyze patterns, create visualizations, and record evidence.

The key part of our build was integrating WebMCP tools directly into the page. Observatory registers tools that an AI agent can discover and use, allowing actions such as starting an investigation, querying data, comparing groups, creating charts, and recording insights. The investigation process is visible through the activity timeline and Investigation Graph, making the agent’s actions easier to understand and review.

We also focused on making the experience interactive and visual. The platform includes data-quality information, suggested questions, 3D visualizations, evidence tracking, investigation history, and a final investigation summary.

The result is Observatory: a workspace where AI does not just tell users what it found, but visibly investigates the data and builds an evidence-based result step by step.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Observatory

Built With

• ai
• csv
• dataanalysis
• javascript
• webmcp