Cotext.ai

Inspiration

We were tired of getting buried in texts, losing key info in messy group threads, and jumping between apps just to answer simple questions.

What it does

Cotext is your unified messaging thread for life. It lets you talk to all your systems in one place — while building and sharing no-code personal agents with scoped access for your friends, family, coworkers, and even the public. Your friends and family can text your agent and it answers questions for you, based on your connected data (scoped to friend or family groups). Example use case: when your family member needs something like your frequent flier number, but you’re away from your phone, then your agent can respond for you and provide that number, based on scoped Notion access.

How we built it

IDEs: Vercel v0.dev for front-end (fast iteration and a clean mobile-friendly UI) and Cursor for the API integration
LLM: Powered by Anthropic’s Claude 3 Opus for test responses and fallback logic
Orchestration & Back-end Logic: deployed multiple MCP clients and servers into our agent orchestrator built in Langflow.
MCP Integrations: Notion, Discord, Gmail, Google Calendar, Outlook, Google Contacts
Frontend: Next.js 15, React 19, Tailwind CSS

Challenges we ran into

Switching late to the LangFlow API set us back – starting here first would have allowed us to iteratively add more MCPs
Mapping from our test scaffolding to real MCP contexts was harder than expected and slowed us down
auth between our MCP servers and 3rd party APIs was finicky and took iteration.

Accomplishments that we're proud of

Fully built each layer of the stack: frontend, agent logic, and working MCP integrations
The app experience feels smooth and already hints at real productivity value
Got some of the MCP tools working in production, but not all of them

What we learned

Which architecture and tools helped us ship quick, and where we lost time
Deepened our understanding of how MCP works, especially the client/server auth model and orchestration patterns

What's next for Cotext.ai

We want to finish a stable production-ready MVP, tune up the app and orchestrator integration, polish the UX, ensure all core MCP tools are functional and deploy an MVP to share with friends.