Tested technical tutorials, generated on your infrastructureautomatically.

An AWS account with permission to provision EC2 (or Fargate), RDS PostgreSQL with pgvector for RAG, and S3. The engine itself is built on Windmill (open-source workflow engine) plus a custom orchestration layer — both run as containers in your VPC. Typical Starter footprint: one t3.medium EC2 + a db.t4g.small RDS + a few GB of S3 — roughly $95/mo on AWS. Growth (~$350/mo) adds Multi-AZ, ElastiCache, ALB, and NAT GW for production-grade workloads. Scale (from ~$1,200/mo) is ECS Fargate with Multi-AZ NAT, CloudFront, and HA across regions. Numbers include ~25% buffer over published on-demand rates to absorb data transfer, snapshots, and CloudWatch ingestion — typical bills come in at or below quoted. You receive Terraform modules so the deployment is reproducible and auditable.

Starter: 1–2 weeks from kickoff to first generated tutorial. Growth: 3–4 weeks (includes fine-tuning on your docs). Scale: 4–6 weeks (includes HA setup and custom integrations). The bulk of the time is RAG indexing, voice calibration, and wiring your code-execution sandbox to your stack — not raw infra provisioning.

Two layers. First, RAG: we index your docs into a vector store so the engine retrieves authoritative context for every tutorial it writes — no hallucinated APIs. Second (Growth and Scale): we fine-tune a voice model on your existing published content so output matches your house style. Both layers stay in your VPC. We never copy your docs out of your infrastructure.

Every code block in every tutorial is executed in a sandboxed CI step before the tutorial is marked as ready to publish. If the code fails, the engine regenerates the block (with the error as context) until it passes — or the tutorial fails CI and lands in a human review queue. Tutorials that ship have code that runs against the dependency versions and APIs you configured. No broken examples.

You do, directly to the provider. We use your API key — no markup, full cost visibility on your provider's dashboard. Typical token cost is $2–10 per tutorial: short Sonnet-only runs land near the low end, while longer tutorials with multiple code-test regeneration loops push toward the upper end. Most customers average ~$5/tutorial on a mixed Sonnet + Opus pipeline. You can switch providers (or use multiple) without touching the engine.

No, by default. The engine runs in your AWS account, in your VPC. Your documentation, your API key, and your generated content stay in your infrastructure. We connect only for the initial deployment and for updates you authorize — typically via a short-lived IAM role you can revoke at any time. Nothing is exfiltrated.

Marginal cost per tutorial is essentially the LLM API call — usually $2–10 in tokens (averaging ~$5) depending on length and which model you point it at. The AWS infra is a fixed monthly footprint ($95–$1,200/mo depending on tier) that you can resize on AWS anytime — tier capacity numbers are sizing guidance, not enforced caps. Comparable agency cost is $300–1,500 per article — so on Growth, fifteen generated tutorials roughly cover a year of infra.

You keep the deployed engine. It runs on your AWS, with your API key — there is no kill-switch on our side. What stops is our ongoing maintenance, model updates, and support. You can self-maintain, hire your own team to maintain it, or re-engage us later. No lock-in.

fastDraft is built by Anchorly LTD, the company you contract with. Our team has 75+ technical articles published across the dev-tool ecosystem (Supabase, Snyk, Redpanda, Mailgun, Merge, Dropbox, Gusto, Stytch, and others) — that's the editorial bar the engine was trained to hold. We've been hand-writing technical content for dev-tool companies for years; the engine is the operating manual for that craft, productised.