Eight apps on one dataset.
Underneath the operator is a real platform. One shared dataset, one AI that reads and writes across all of it, and one meter. Here's what it's built on, for the person who wants to look under the hood before they trust it.
One data model, not eight silos
The eight apps — Website, Email, Forms, Knowledge Base, CRM, Chat, Developers, and Socials — are not eight products bolted together. They read and write the same Postgres dataset. What a form captures is available in the CRM. What the AI learns in chat is available when it writes an email. Nothing gets copied between tools, and nothing falls out of sync.
The whole thing is a polyglot monorepo: TypeScript for the API, web app, website frontend, and the AI service; Python for content and file processing. Fastify with TypeBox schemas generates the OpenAPI spec; the frontend is React with TanStack Query and Router. It all runs on Google Cloud — Cloud Run for compute, Cloud SQL running PostgreSQL with pgvector for embedding search, Cloud Storage for files.
The surface, in numbers
0
Integrated apps
0
Tools the AI can call
0
Content blocks
0
Field types
The stack
Production systems, not prototypes. Thousands of automated tests run on every push; the release pipeline promotes to production only when the full suite is green.
The AI service, and how you reach it
Content generation runs on Google's Gemini family today — writing, embeddings, and image generation — with the model chosen per job. A cheap model classifies; a stronger one generates; a vision model checks the rendered page. Every call routes through a provider-agnostic layer that can adopt other models by config as they become viable, so the platform isn't welded to one model.
You talk to it in the LLM client you already use. Giant Context is MCP-native from day one — the AI receives instructions through Claude or Cursor, not a dashboard you have to log into. There's a web console if you want one, but most days you won't touch it. Machine access is through API keys, scoped per organization and revocable at any time.
How the generation pipeline actually grounds a draft — facts versus context, the quality gates, the vision check — has its own page.
One meter, one rate
The meter runs with the AI. When the AI stops, the meter stops. You start capped at $50, and the console shows every operation and what it cost.
Your data, and the honest state of things
Your data is yours. It lives in your project on Google Cloud, in US regions today, with encryption in transit and at rest. It does not train any public model, and it does not leak into another customer's context. You can export everything, in bulk, at any time. If you leave, you take it.
And the honest part: this is early alpha, built by one person. There's no SOC 2 yet, no SSO on the base plan, and no formal SLA — all on the roadmap, none shipped. EU data residency is an enterprise option that isn't built. If you need those today, this isn't your tool today, and we'd rather say so than pretend.
Connect your LLM client and go
Early access grants sandbox access to the whole platform. Upload your business docs first — the AI grounds everything it produces in what you give it — and start asking. If you want to read the code-level detail before you sign up, ask in chat.