Architecture Workbench · Diagrams · Decisions · Costs

Where Architecture Decisions Become Executable Systems.

In most teams, architecture lives in three places that never agree: diagrams, decision documents nobody updates, and a cost spreadsheet full of guesses. ArchiMesh keeps all three in one shared model that architects, developers and Platform SREs work from together, so the diagram, the decision and the cost always describe the same system.

  • 106 Cloud & Open-Source Building Blocks
  • Diagrams at 4 Levels of Detail
  • Exact Costs, Never Guessed
  • 460+ Automated Tests

The Real Workbench

Live Architecture Telemetry

A working miniature of the online checkout sample (SOL-001) from the live demo. Click the parts of the system, see whether each cloud meets the targets, and recalculate the monthly cost for each environment.

Live Architecture TelemetrySOL-001
Design Score: 92/100
System Boundary · Omnichannel Checkout mTLS SQL/TLS RESP AsyncAPI Edge WAF[Container]Checkout / Core API[Container]PostgreSQL 17 HA[Database]Valkey Cache[Cache]Redpanda / Kafka Outbox[Event Bus]
Checkout / Core API

Runs the checkout on Kubernetes across three data-centre zones. It saves each order together with its outgoing message, so neither can be lost.

The Problem

No More Outdated Diagrams or Patchwork Cloud Stacks

Three problems every architecture team knows, and how ArchiMesh removes each one by design.

1. Diagrams That Stay Current

The Usual Way

Diagrams, decision records and wiki pages are edited separately, so within a quarter they describe three different systems.

With ArchiMesh

Decision records, diagram code and the visual canvas are three views of one model. A change in one appears in the others, and real conflicts are flagged for review instead of being silently overwritten.

2. One Cloud at a Time, by Design

The Usual Way

Stacks built piece by piece end up spread across clouds: a database here, a queue there, and data-transfer bills nobody planned.

With ArchiMesh

Each solution stays on one provider (AWS, Azure, GCP or open source), and one click moves every layer to its match on another. A hybrid setup is still possible, but only as a deliberate choice.

3. Exact Math, Not AI Guesswork

The Usual Way

Costs and rankings come from spreadsheets or chat assistants: rounded, hard to repeat, and blank cells that count as zero.

With ArchiMesh

Costs and rankings are calculated by exact, rule-based code, never by AI. A missing price is flagged until someone enters it with its source; it never shows as $0.00. Same inputs, same answer, every time.

6 Core Tools

Built to Be Used, Not Just Drawn

Each card below is a working miniature of the real tool: switch cloud, change the level of detail of a diagram, apply a suggested fix and compare the generated code.

Two Ways to Start

Start From Requirements, or From a Pattern

Path A Guided Advisor

Enter your targets for uptime, speed, recovery, compliance and budget. The advisor drops the options that miss them, ranks the rest, and a person confirms the choice in a decision record.

Path B Blueprint Studio

The order and its message are saved together and the message is sent afterwards, so nothing is lost or sent twice.

Cloud Switch

Move the Whole Stack in One Click

LayerComponent$/moLock-In
Front Door & Firewall
App Servers
Main Database
Fast Cache
Messaging
Total

Illustrative catalog prices in USD per month, production sizing. Lock-in is the average across layers, out of 100: lower is easier to move.

Diagrams & Design Review

Write the Model, Check It, Fix It

checkout = softwareSystem 'Omnichannel Checkout'
shopper  = person 'Shopper'
psp      = softwareSystem 'Payment Provider' { #external }
shopper -> checkout 'Places orders' 'HTTPS'
checkout -> psp 'Authorizes payments' 'mTLS'

Who uses the system and which external systems it depends on.

Design Score 92/100
  • Single point of failure: the cache runs on one server with no backup.
  • The payment call has no protection if the payment provider slows down.
Portfolio Cost Roll-Up

Three Solutions, One Portfolio Budget

PRJ-COMMERCE

Cloud & Open-Source Library · 106 Entries

Learn It or Use It, Same Entry

AWSAurora PostgreSQL 17 · Multi-AZ
Key Capabilities
PostgreSQL 17 compatible, data copied across three zones, up to 15 read copies, restore to any point in time.
Target Use Cases
Main business databases, order and payment flows, software platforms that serve many customers.
When to Avoid
Heavy analytics over billions of rows, or when you must stay fully portable and a managed service would tie you to one cloud.
Portable Code Export

The Model Exports to Code You Own

module "checkout_eks" {
  source          = "./modules/eks"
  cluster_version = "1.31"
  azs             = ["eu-central-1a", "eu-central-1b", "eu-central-1c"]
}

resource "aws_rds_cluster" "orders" {
  engine                  = "aurora-postgresql"
  engine_version          = "17.2"
  storage_encrypted       = true
  backup_retention_period = 14
}

Team Hand-Off

Design → Build → Deploy, With Checks in Between

One solution passes through three roles. Each sees only what it needs and hands over only when its checks are green. Pick a role to see its stage.

Stage 1 · DESIGN

The architect turns business goals into clear targets and a design the other two roles can trust.

  • Targets for uptime, speed, recovery, budget and compliance
  • Recommended patterns for each target (database copies, reliable messaging, failure isolation)
  • Diagrams at all four levels, with no critical review findings
  • A linked decision record

Check: a design at least 80% complete unlocks “Publish Design to Build →”

From the Portfolio

FinPulse, Architected Inside ArchiMesh

ArchiMesh's main sample is a real system I built: FinPulse, my behavioral finance lab, modeled as three solutions from its own code and decision records.

SOL-FP-01

Financial Records, Account Sync & Analytics

A time-series database for transactions, a fast cache that refreshes on every change, account and exchange-rate sync in its own isolated service, and a separate in-memory engine for analysis.

SOL-FP-02

Private, Local AI Advisor

AI runs first on a dedicated machine on the local network, with read-only access to the data, and every number is re-checked by code before an answer is shown.

SOL-FP-03

One Interface for Dark & Light, One Chart Engine

A modern web app with one shared chart engine, linked charts, Simple and Pro views, and a read-only public demo built from the same release.

PRJ-FINPULSE Explore the FinPulse Case Study Behavioral analytics, private AI and the full system design.

System Architecture

How ArchiMesh Is Built

Pick a flow to watch a request travel through the system, or click any part to see what it does.

Interactive System Graph Self-hosted · Private network · Exports you own
Selected: API server
Details

Quality & Verification

The Engineering Quality Scorecard

How the codebase is kept honest: twelve recorded design decisions settle the hard choices up front, and every release passes the checks below.

1 Query

Fast Hierarchy Queries

Parent-and-child links are worked out in advance, so questions about what inherits from what take one simple query. Tests run the same database in memory.

100%

Checked Data at Every Entry

Every request to the server is checked against a schema, and the same schema defines the types in the code, so the two cannot drift apart.

460+

Automated Tests

Test suites for the server, the web app and every engine, each required to cover at least 80% of its code.

469

Components Checked Before Release

A full inventory of the 469 software components is checked offline against an open-source licence policy before every release.

WCAG 2.1 AA

Accessible in Dark & Light

Both themes meet AA contrast, every control works from the keyboard, and phones get their own bottom tab bar.

Open the Workbench and Explore

The public demo is read-only: browse the two sample projects (Commerce and FinPulse), run the advisor, open the decision records and check the costs. Editing is reserved for signed-in workspaces.

Self-Hosted on Open Source · Private Encrypted Network · No Usage Tracking

Architecture & advisory

Planning a cloud platform, a migration, or an architecture practice?