Thothstack

• 

  Thothstack

• 

  Logo

Inspiration

We were inspired by the growing complexity of managing LLM-based agent systems and the friction that researchers and devops teams face when turning AI logic into cloud-native microservices. We wanted a platform where agent deployment could be as simple as writing a config file.

What it does

ThothStack generates and deploys multi-agent systems using GitLab and Google Cloud. It auto-generates agents, CI pipelines, cloud infra, and exposes secure public endpoints. It’s like Vercel for AI agents.

How we built it

We built a TypeScript + React front-end with Tailwind UI to guide users through an agent configuration wizard. The backend runs on Netlify Functions to package and upload code. GitLab CI/CD handles build and deploy stages. Images are pushed to Artifact Registry, and Cloud Run hosts the containerized agents.

Challenges we ran into

Google Cloud IAM and service account permissions Docker image push errors to gcr.io and transitioning to Artifact Registry Making the CI/CD generic across all agent types while allowing customization GitLab runner environments lacking bash/curl/python by default

Accomplishments that we're proud of

Our build worked end-to-end in the hackathon environment Multi-tenant based AI Agent deployment framework Prompt-to-Agent Generation with realtime AI agent deployment AI Agent templating with standard config

What we learned

Cloud Build and Artifact Registry are powerful but require strict permission boundaries GitLab CI/CD can be templated elegantly for agent pipelines A prompt-driven UX dramatically reduces agent setup time

What's next for Thothstack

Integrate with LangChain and HuggingFace for agent tooling Integrate with AI agent Hosting platform Persistent vector memory storage Visual pipeline editor and deploy dashboards Team-based multi-agent deployment and monitoring

Built With

• gcp
• gitlab
• javascript
• python
• typescript