WriterzRoom: Governed multi-agent AI content for regulated industries

100% serverless on Google Cloud: 9 Cloud Run services (4 product, 5 internal ops agents), zero VMs to manage.

Hey Product Hunt, I’m Jason, maker of WriterzRoom.

Generic AI writers produce text. Regulated teams need drafts a reviewer can sign off on. WriterzRoom runs each request through an 8-agent LangGraph pipeline, constrained by template schema and vertical controls, and held at a mandatory Quality Gate until it passes. Professional and Enterprise plans add a claim-by-claim evidence record.

Our checks are deterministic rules, and a suite of documents with planted defects runs on every pull request: 23 of 23 seeded defects caught, 0 of 13 clean documents blocked. You can check a draft free, no credits needed.

The video above is an animated walkthrough with illustrative data.

How we use Cloud Run:

writerzroom-backend-api: FastAPI + LangGraph orchestration. Scales with generation and publishing load, and Cloud Run’s request timeouts handle long-running agent chains.
writerzroom-frontend-web: Next.js 15 (1 vCPU / 1 GiB, 1→5 instances, health-checked startup probe), serving writerzroom.com.
writerzroom-mcp and writerzroom-corpus-mcp: two MCP servers that expose platform tools and the content corpus to the agents.
Internal ops, same stack: five agents (analytics, competitive intel, prospect research, outreach, content pipeline) that I use to run WriterzRoom itself, deployed the same way as the product.
Wiring: Cloud Build CI/CD (a regression suite gates every deploy) → Artifact Registry images → Cloud Run. Secret Manager for every credential, a Serverless VPC connector to private Cloud SQL Postgres, per-service IAM service accounts, and Cloud Scheduler for auto-publishing.

Why Cloud Run: elastic under load, one deploy path for every service, and no servers to patch.

Happy to dig into any of it: how we run LangGraph chains on Cloud Run, the validation layer, or how the services stay wired together. What would you want to see next?