• 

  Logo

Campion

Inspiration

Finding a good dispersed campsite often means opening dozens of tabs, cross-referencing maps, digging through Reddit threads, and trying to match all of it to personal preferences like accessibility, scenery, remoteness, and nearby amenities. What should feel exciting turns into hours of scattered research. We wanted to build something that makes this process faster, smarter, and more personalized, so you spend less time researching and more time outside.

What It Does

Campion helps users discover dispersed camping locations tailored to their preferences. Instead of relying on a single source, our system deploys multiple specialized Browser Use agents that search across topographic data, Reddit discussions, camping websites, and other online resources in parallel. The agents gather, compare, and filter information to surface spots that best match what the user wants - including locations that have never been reported online, identified through terrain and environmental analysis. Results are displayed on an integrated map so users can visually explore and compare locations. Campion turns hours of camping research into a couple of minutes.

How We Built It

Campion is a multi-agent system built on a FastAPI backend with a single-page HTML/CSS/JS frontend. Each Browser Use cloud agent has a focused responsibility - one agent explores topographic and terrain data, another pulls community advice from Reddit, and others aggregate camping resources across the web. Agents run in parallel, and their findings are combined into a unified recommendation flow based on the user's selected preferences. Results are surfaced through an integrated map interface so users can immediately see where recommended spots are located. The architecture was designed around one principle: break a complex research task into smaller, specialized tasks and let agents handle them simultaneously.

Challenges We Ran Into

Coordinating multiple agents was the biggest technical challenge. Early on, agents would open but not execute actions, which required careful tuning of task prompt length and run timeouts to get reliable behavior. We also dealt with messy, inconsistent information across sources - camping advice online is often subjective, outdated, or incomplete - making it difficult to compare results reliably. Translating user preferences into search behavior that actually produces useful campsite recommendations required more iteration than expected.

Accomplishments We're Proud Of

We built a system that approaches camping research the way a real human researcher would - investigating multiple angles of the same problem simultaneously - except much faster. We're also proud that Campion surfaces genuinely undiscovered spots using topographic and environmental analysis, not just aggregating what's already online. Most importantly, it solves a real pain point and delivers personalized results in a visual, accessible format.

What We Learned

Multi-agent workflows are especially powerful when a task requires pulling information from many sources and synthesizing it into a single answer. We learned that browser-based agents become significantly more effective when each one has a clear, scoped role. On the technical side, prompt design and timeout management turned out to be critical factors in agent reliability. Beyond that, we learned how important it is to design around the user's actual decision-making process, not just around what data is easiest to collect.

What's Next for Campion

We want to improve how Campion ranks and verifies campsites, expand the preference system for more granular user input, and add better explanations for why each location was recommended. On the map side, we're planning to add route planning, nearby points of interest, and stronger trust signals so users can quickly evaluate recommendations. Long term, our goal is to evolve Campion into a full outdoor planning assistant - not just a campsite search tool.

Built With

• browser-use
• css
• javascript
• mapbox
• python