Architecture templates with the cost already on them.
22 systems you can open on the canvas in one click — cloud infrastructure, AI agents, and the ones that are both. Every template carries a projected monthly cost, so you know what a design is worth before you build it.
Support agent on live systems
A support agent that reads the orders database and writes back through an approval gate. Agent steps and the systems they touch on one graph.
Three-tier web application
The default shape: load balancer, containerised app tier, managed Postgres, cache and asset storage.
Agent with enforced write control
The pattern for letting an agent touch production: every write passes an approval gate, every action lands in a hash-chained evidence log.
Serverless API
HTTP API with function compute and a key-value store. Scales to zero; the floor is the gateway.
Retrieval service (RAG)
An answer endpoint over your own documents: vector store, retrieval step, generation step, cached responses.
Multi-account landing zone
Separate accounts per environment, one org-wide audit trail, archived logs and a budget guardrail on top.
Static site on CDN
Marketing site or SPA served from object storage behind a CDN, with DNS and a WAF in front.
Three-tier app in a VPC
The same three-tier shape drawn the way AWS actually organises it: edge in a public subnet, app and data in a private one, and the pieces that live outside the VPC outside the box.
Container microservices
Several services behind one load balancer, sharing a database, a cache and an internal queue.
Internal tools behind VPN
Admin and back-office apps reachable only over the VPN, on a single instance with a small database.
Nightly batch ETL
Scheduled extract and load from source systems into a lake, then modelled into a warehouse.
Streaming ingest
Events land on a stream, are enriched in flight, and are written to both a lake and a search index.
Analytics warehouse
Lake, warehouse and BI, with archival storage for anything older than the retention window.
Queue-backed workers
Accept work fast, process it out of band, and keep failures in a dead-letter queue instead of the logs.
Event fan-out
One published event, several independent consumers, each with its own queue and failure path.
Document processing pipeline
Files land in storage, an agent extracts structured fields, a person approves anything low-confidence, and the result is written to the database.
Bulk classification
High-volume, low-cost classification over a queue, with the cheap model doing the work and a cap on the bill.
Vercel + Supabase app
The default modern startup stack: hosted front end, managed Postgres and auth, edge cache, payments.
Fly.io app and database
Machines close to users, Postgres in the same region, Redis for sessions and rate limits.
Railway monolith and worker
One web service, one background worker, a serverless Postgres and object storage. The smallest real stack.
GPU batch on Modal
Bursty GPU work triggered from a queue, with inputs and outputs in cheap object storage.
Multi-region active-passive
Primary region serving traffic, second region warm with a replicated database and cross-region backups.
Costs are projections from August 2026 list prices, assuming always-on resources. Connect a cloud account and the same nodes carry your billed figures instead.
Open any of them on the canvas.
Templates load as real, editable graphs — the same ones the copilot and the runtime read.