Inspiration

Today, 30% of college students are considered clinically depressed. 60% face dangerous amounts of anxiety and stress.

As students in a demanding dual-degree program, we often push mental health aside, telling ourselves "I’ll be fine, I just need to push through." But therapy access is full of friction: finding providers, scheduling appointments, and dealing with stigma around asking for help. We wanted to flip the model from reactive crisis response to proactive prevention - like a friend who "checks" (chex) in on you, handles logistics, and finds you help before things break.

Additionally, having worked in the mental health field, we've noticed that a large problem therapists face nowadays is the lack of out-of-clinic info regarding their patients.

Chex solves both problems at once, removing barriers to support and enabling therapists to better serve their patients

What it does

Chex proactively deploys three agents.

1. Performs empathetic check-ins

   • Aggregates student context (email, calendar, journal) via Arcade integrations with Google Calendar, and Notion to detect stressful periods.
   • Calls students using Vapi’s warm, customizable, and HIPAA-compliant voice AI to hold real, human-like conversations.
   • The idea is not to be a therapist, but to be a "friend" who checks on you and analyzes your responses to determine if you need help.
   • Chex automatically takes action based on the call transcript. For example, if we notice that the student expresses that their life is falling apart, we find them a therapist - then share the patient's history with their current or new therapist to provide the providers with the context they need to provide the best care possible.

2. Finds therapists & resources

   • Uses BrightData MCP to find relevant mental health resources and therapist names (campus + local), since chex has context about the student's university and location.
   • Uses BrightData MCP + Firecrawl (Arcade) to look up and enrich therapist profiles we found with phone, email, address, and specialties.

3. Schedules appointments automatically

   • Uses Vapi to automatically call therapist offices on behalf of the student, confirm availability, and check against the student’s calendar, giving the student only options that work for them.
   • Books appointments via Mastra workflows, again using our integrations and toolkits via Arcade to sync calendars and emailing confirmations including coping resources.

How we built it

We built three specialized AI agents orchestrated by Mastra:

• Check-In Agent: proactive Vapi voice calls and sentiment-driven triage, gathering data about the student's life and mental state.
• Therapist & Resource Agent: BrightData MCP for fast lookups + Firecrawl + Google Search for detailed therapist profiles.
• Scheduling Agent: automated booking, calendar sync, resource delivery to the student, and patient history sharing with the therapist.

Challenges we ran into

• Normalizing multiple data sources into one schema (Arcade).
• Designing empathetic voice experiences (Vapi).
• Matching therapist availability to packed student schedules.
• Orchestrating five steps into three robust agents without latency (Mastra).

Accomplishments that we're proud of

• Multi-source context aggregation via Arcade (Notion journals, Gmail, Calendar).
• Voice empathy engine via Vapi (HIPAA-compliant, customizable voices).
• Real-time resource discovery via BrightData MCP.
• Web scraping and data enrichment using Firecrawl + Google Search (Arcade).
• Orchestration and automation via Mastra, chaining all agents seamlessly.

What we learned

Working with several sponsors' tech and integrations was a great learning experience - what we previously used to spend hours on, we no longer have to "reinvent the wheel" for.

• We've worked with Vapi before, but a lot of great recent features really helped us. For example, last month's release of "Workflows" in Vapi enabled us to build a more empathetic, smooth, and human-like experience with real agentic capabilities.
• It was our first time working with Mastra, and we were blown away by how powerful it can be, as a TypeScript-native workflow orchestration engine.
• Arcade: With integrations across various platforms being such a pain, finding a "hub" like Arcade was a game-changer. We saved a lot of time, gathered a lot more data, and were able to build a more robust product. Definitely something we will use in the future.
• BrightData MCP: We experimented with it at yesterday's hackathon, but didn't end up using it. Today, we got that chance, and were impressed with how easy it is to gather real-time data without facing obstacles like rate limits, CAPTCHA, etc.

What's next for chex

Chex is poised to make a significant impact by lowering barriers to care and building trust through proactive support:

• For students: Faster, stigma-free access to help.
• For therapists: Clients arrive prepared, reducing scheduling overhead.
• For universities & insurers: Preventative mental health support lowers long-term costs and risk.

Commercial Potential:

• University contracts (per-student, per-semester pricing, similar to Calm or Headspace)
• Insurance partnerships (lower claim costs via early intervention)
• Direct-to-therapist & direct-to-patient models (streamlined booking and engagement)
• Corporate wellness expansion (future pivot to employee assistance programs)

Planned Next Steps:

• Wearable integration (Fitbit, Apple Watch) to detect physiological stress
• HIPAA-compliant EHR integrations for provider data sharing
• Mobile-first experience for students on the go
• Expansion to high-stress populations beyond students (corporate, healthcare)