About BackArch · Feature catalog

We believe great software
starts with great architecture.

BackArch is the backend architecture platform for engineers who take their systems seriously — five design surfaces, a constraint-aware AI co-pilot, a load and failure simulator, live multi-cloud pricing, and a deterministic health engine in one workspace. Everything below is shipped and usable today.

0
Shipped features
5
Design surfaces
0
Catalog components
0k+
Live pricing records
32
Patterns with demos
22
Health rules

Our mission

Make world-class backend architecture accessible to every engineering team — not just the ones with a principal architect on staff. The knowledge, the guardrails and the tooling, in one place.

🔭

Our vision

Every deployment decision made with full awareness of its cost, its resilience and its architectural merit — captured in a living document the whole team can read, evolve and trust.

🏗

Who we build for

Backend engineers designing new systems or evolving old ones. Platform teams who own the infrastructure decisions. Tech leads who need architecture to be legible outside one person's head.

Design Surfaces

5 features

Five editors that cover architecture, behaviour, conversation, schema, and the writing around them.

01·Design SurfacesFree + Pro

Architecture Canvas

The whiteboard that knows what it's drawing.

464
catalog components
22
service categories
6
connection types
4
canvas tools

Drag real infrastructure onto an infinite canvas and connect it with typed edges. Every node maps to a billable cloud service, so cost and health update as you draw.

  • AWS, GCP, Azure and cloud-neutral components, all searchable by category
  • Typed edges — sync, async, stream, bidirectional, dependency, data flow — each carrying auth, encryption, DLQ, retry, timeout and SLO metadata
  • Instance picker and t-shirt sizing on 200+ components, feeding the cost estimate live
  • Dockable, floatable, resizable panel rails for properties, AI, and health
  • Select / pan / connect / text tools, multi-select, group actions, minimap
  • Auto-save on a 2.5s debounce, then publish when you're ready to freeze a version
Auto-saveMinimapPNG exportMarkdown exportStar & tagDuplicateFind it atSidebar → Architecture
02·Design SurfacesFree + Pro

Sequence Diagram Editor

Make every service-to-service conversation legible.

5
participant kinds
6
message kinds
6
fragment types
3
export targets

Full UML sequence semantics on a canvas that stays readable past eight lifelines. Draw the call order, then export it into your README or design doc.

  • Actors, services, databases, queues and external systems, each with its own lifeline styling
  • Sync, async, return, self, create and destroy messages with correct arrowheads
  • alt / opt / loop / par / break / critical fragments, nestable, preserved on export
  • Step or auto-play playback that walks the conversation message by message
  • Drag to reorder participants; columns snap to a grid
  • AI panel sees every lifeline and message and can add error branches or fan-outs
MermaidPlantUMLPNGPlaybackAI assistFind it atSidebar → Sequence
03·Design SurfacesFree + Pro

State Machine Designer

Model behaviour, not just boxes.

7
state kinds
3
action slots
4
export targets
1
built-in simulator

The workflows and lifecycles that boxes-and-arrows diagrams never capture. Draw the states, step through the runs, then export straight into XState.

  • Initial, normal, final, choice, fork, join and history pseudostates with UML structural rules enforced
  • Transitions carry a triggering event, a guard expression and an action, edited inline
  • Entry / do / exit actions per state, fired visibly during simulation
  • Simulator steps events one at a time, highlighting the active state and logging side effects
  • Validate panel flags unreachable states, missing initial transitions, dangling forks, guardless choices
  • AI panel can add retry loops, generate guards, or explain why a state is unreachable
MermaidPlantUMLXStatePNGSimulatorValidatorFind it atSidebar → State Machine
04·Design SurfacesFree + Pro

Data Model Designer

Design the schema. Ship the migration.

6
entity kinds
9
relationship kinds
7
code targets
2
view modes

An entity-relationship canvas that speaks both DBA and engineer. Same model, two views, seven code targets.

  • Tables, views, enums, interfaces, abstract and concrete classes, each with its own header style
  • Typed fields with nullability, defaults, and index / unique / primary-key flags
  • Method signatures with parameters, return types and visibility on class-kind entities
  • Association, 1:1, 1:N, N:M, inheritance, composition, aggregation, realization, dependency
  • Toggle between Database view (tables, columns, FKs) and Code view (classes, methods)
  • AI panel suggests missing indexes, junction tables, audit columns, or a normalisation pass
PostgreSQLMySQLSQLitePrismaTypeScriptPythonJavaPNGFind it atSidebar → Data Model
05·Design SurfacesFree + Pro

Notes & Docs

The writing that explains the drawing, in the same workspace.

6
heading levels
3
callout styles
/
slash commands
S3
attachments

A block editor for design docs, runbooks and research, organised in folders next to your diagrams. Slash commands, callouts, attachments, AI drafting.

  • Folders with emoji, plus All Notes and Starred views across everything you've written
  • Type / anywhere for headings, tip / warning / info callouts, code blocks, quotes, dividers
  • Auto-generated table of contents that tracks the heading you're reading
  • File attachments stored on S3 and listed per note
  • Tags, sections, starring, and one-click Markdown export
  • AI assist drafts or rewrites a section in place, and every new account starts with a guided welcome note
Markdown exportTOCAttachmentsTagsAI assistFind it atSidebar → Notes

AI & Intelligence

5 features

Everything that reads your diagram and tells you something you didn't already know.

06·AI & IntelligenceFree + Pro

Constraint-First AI Co-pilot

AI that knows your budget, your cloud, and your diagram.

8
canvas tools
1-click
apply to canvas
Live
token streaming
session history

Not a chatbot bolted onto a diagram tool. Every message carries your full canvas state and your constraints, and the answers come back as changes you can apply.

  • Constraint profile — cloud, budget, traffic, compliance, team size — injected into every request
  • Eight callable tools: health score, anti-pattern check, place components, modify diagram, pattern suggest, ADR search, ADR draft, price lookup
  • Structural suggestions arrive with an Apply button that mutates the canvas and snapshots a version
  • Responses stream token by token, with tool calls narrated as they run
  • Conversations are stored per diagram, so you can reopen a thread days later
  • Every editor has its own AI panel grounded in that surface — canvas, sequence, state machine, data model, notes, patterns, pricing
StreamingTool useApply to canvasPer-surface panelsFind it atAny editor → AI panel
07·AI & IntelligencePro

Deep Review

A principal engineer read of the whole diagram.

Sonnet
review model
8
credits per run
Whole
diagram scope
Pro
plan

One command runs a full architecture review on the deep model — it reads every node, edge and finding, then writes up the resilience, security and cost issues a rule engine can't see.

  • Runs on Claude Sonnet 4.6 while everyday chat stays on the fast Haiku tier
  • Reads the deterministic findings first, then reasons about what they imply together
  • Surfaces emergent risks — coupling, blast radius, ordering assumptions — not just rule hits
  • Output is streamed as a structured review you can act on finding by finding
  • Recommended actions resolve to real catalog slugs, so they apply directly to the canvas
Claude Sonnet 4.6StreamingStructured actionsProFind it atCanvas → AI panel → Review
08·AI & IntelligenceFree + Pro

Architecture Health Score

Your architecture gets a grade, and every point is fixable.

22
deterministic rules
6
scored categories
5
readiness tiers
0
guesswork

A 0–100 composite across six weighted categories, computed by a deterministic rule engine — not a vibe. Each deduction names the components responsible and the fix.

  • Reliability 25%, Security 25%, Scalability 20%, Observability 15%, Cost 10%, Maintainability 5%
  • Critical violations cap a category outright — no auth on an external path can't be averaged away
  • Readiness tiers gate on rule IDs: Prototype → Dev → Production → HA → Enterprise, with the exact blockers listed
  • Topology analysis computes structural SPOFs, blast radius per node, critical sync-path length, and fan-in / fan-out
  • Findings carry severity, penalty, affected nodes and edges, a written fix, and a link to the matching pattern
  • AI observations run after publish to name structural risks the rules can't express
SPOFDLQCircuit breakersAuth & TLSTracingBudgetCouplingFind it atCanvas → health ring in the top bar
09·AI & IntelligenceFree + Pro

Generate & Scan-Import

Start from your real system, not a blank canvas.

6
context questions
3
starting paths
Local
scan stays local
Pro
for scan-import

Answer six questions about your stack, then either let AI scaffold the diagram or scan the system you already run and import what comes back.

  • System context captures cloud, system type, IaC, use case, scale, and what you optimise for
  • Generate builds a fully wired, health-scored diagram from that context plus a description
  • Scan hands you a prompt to run against your own codebase with your own tooling — nothing is uploaded
  • Paste the scan output back and BackArch reconstructs the topology as an editable diagram
  • Or start empty and drag from the palette; the context still shapes every AI answer afterwards
BlankGenerateScan-importContext-awareFind it atDashboard → New diagram
10·AI & IntelligenceFree + Pro

AI Model Controls & Credits

Pick the engine. Watch the meter.

3
model tiers
8
metered actions
Per-task
overrides
Monthly
credit reset

Choose the model tier behind your co-pilot globally or per task, and see exactly what each AI action costs before you spend it.

  • Fast and Balanced run on Claude Haiku 4.5; Deep runs on Claude Sonnet 4.6
  • Set a default tier, then override individual tasks — classification, summaries, architecture, health, ADRs, review
  • Credit weights are published up front: chat 1, notes 2, generate 3, scan-import 5, deep review 8
  • Usage table breaks the billing period down by action, call count and model
  • Running out pauses AI only — canvas, health, pricing and diagrams keep working
FastBalancedDeepUsage breakdownFind it atSidebar → AI Models

Simulation

1 feature

Run traffic and failures through the design before anything is provisioned.

11·SimulationFree + Pro

Load & Failure Simulation

Break it here, not in production.

7
scenario presets
5
traffic patterns
4
failure modes
1s
simulation ticks

Push traffic and failures through the diagram and watch it respond tick by tick — queues filling, autoscalers reacting, SLOs breaking. Then get the ranked list of what to fix first.

  • Presets: steady state, Black Friday spike, viral launch ramp, daily cycle, database outage, region outage, cold cache
  • Traffic dial for baseline RPS, peak multiplier, spike window, payload size, read/write split, cache hit rate
  • Inject failures per node — down, degraded, slow, partition — with start time, duration and severity
  • Model retries, backoff, timeouts and circuit breakers, plus single / active-passive / active-active regions
  • Timeline scrubs the whole run with a p95 chart, SLO breach bands and autoscaling markers; particles and a load ribbon show flow on the canvas itself
  • Results rail returns max sustainable RPS, SLO verdicts, ranked bottlenecks with one-click fix patches, a capacity runway table, and cost now / at peak / at horizon
Max sustainable RPSBottleneck rankingCapacity runwayCost projectionSLO verdictsSaved scenariosFind it atCanvas → Simulate mode

Cost Intelligence

2 features

Live cloud and model pricing wired into every component you place.

$
12·Cost IntelligenceFree + Pro

Live Multi-Cloud Pricing

Know the bill before you provision anything.

148k+
pricing records
174
cloud services
31
regions
Weekly
refresh cron

A refreshed pricing table pulled straight from AWS, GCP and Azure, wired into every component on your canvas and browsable on its own.

  • Fetched per provider on a weekly cron with atomic per-provider replacement — a failed fetch keeps the old rows rather than blanking them
  • Freshness banner tells you exactly how current each provider's data is
  • Filter by provider, service, region, instance type and pricing dimension; export the result as CSV
  • Compare view maps nine service groups — compute, managed DB, cache, API gateway, object storage, serverless, Kubernetes, NoSQL, networking — across all three clouds
  • Eleven geo-region groupings so you compare US East to US East, not to whatever the provider called it
  • Per-component cost lands in the properties panel and rolls into a diagram total; the AI panel answers pricing questions against the same table
AWSGCPAzureCSV exportCheapest lookupCompare viewFind it atSidebar → Pricing → Infra
13·Cost IntelligenceFree + Pro

AI & LLM Model Pricing

Inference is infrastructure. Budget it like infrastructure.

10
model providers
13+
cost types
per 1M
normalised tokens
Live
cheapest cards

Token pricing for every major model provider in one sortable table, so an architecture with a RAG pipeline gets a real cost number instead of a shrug.

  • Anthropic, OpenAI, Google, DeepSeek, Mistral, Groq, Cohere, xAI, AWS Bedrock, Azure OpenAI
  • Input, output, cached reads and writes, batch, audio, image and search costs — all normalised per million tokens
  • Multi-select provider and cost-type filters with live counts, plus model search
  • Cheapest-per-model cards surface the price floor at a glance
  • Stat row shows models matched, providers covered, cheapest input, and the spread across the set
Sortable tableProvider filtersCheapest cardsCost spreadFind it atSidebar → Pricing → AI Models

Knowledge Library

3 features

Patterns, production architectures, and reading — all interactive, all in-app.

14·Knowledge LibraryFree + Pro

Pattern Library

32 patterns, each with a demo you can actually play with.

32
patterns
32
interactive demos
10
categories
2
code languages

Systems and AI patterns written out properly — roles, tradeoffs, pitfalls, code in two languages — and every single one ships with an interactive demo.

  • 22 systems patterns across behavioural, resilience, architectural, creational and infrastructure
  • 10 AI patterns across prompting, agentic, reliability, evaluation and integration — RAG, ReAct, tool use, guardrails, semantic cache and more
  • Every pattern: overview, plain-language analogy, participant roles, when to use, when to avoid
  • Real-world examples naming the systems that use it, plus the pitfalls that bite people
  • Runnable TypeScript and Python side by side, stored as content so they update without a deploy
  • Beginner / intermediate / advanced difficulty, related-pattern links, and AI suggestions for what your diagram is missing
Circuit BreakerCQRSSagaOutboxRAGReActGuardrails+25 moreFind it atSidebar → Patterns
15·Knowledge LibraryFree + Pro

Template Gallery

Production architectures, cloned in seconds.

14
built-in templates
6
real-world case studies
2
template kinds
Fork
or clone

Start from a wired, costed, health-scored diagram instead of an empty grid — including reverse-engineered builds of systems you already use.

  • Case studies: Uber, Netflix, WhatsApp, Twitter/X, BookMyShow, Swiggy — each a full annotated topology
  • Starters: standard web app, microservices, serverless API, event-driven pipeline, ML pipeline, batch + streaming data, notification fan-out, realtime app
  • System templates clone a whole architecture; component templates drop a reusable block into an existing one
  • Every template ships with node and edge counts, an estimated monthly cost, and a health score before you clone
  • Case-study templates come with a pre-seeded AI conversation explaining the design decisions
  • Pro accounts can save their own diagrams and blocks as private templates
UberNetflixWhatsAppTwitter/XBookMyShowSwiggyCustom (Pro)Find it atSidebar → Templates
16·Knowledge LibraryFree + Pro

Learning Hub

A curated reading list that remembers where you left off.

8
topics
7
resource types
%
reading progress
My Zone
personal shelf

Books, articles, videos, courses and tools worth your time, organised by topic — plus your own private shelf.

  • Topics: system design, backend engineering, DevOps & infra, microservices, databases, security, AI/ML infra, career
  • Types: books, articles, videos, courses, tools, newsletters and sites, each badged and levelled
  • Bookmark anything and track reading progress as a percentage per resource
  • My Zone is your own library — add any URL with a title, type, topic and level
  • Beginner / intermediate / advanced filters, plus a topic cloud for browsing sideways
BookmarksProgress trackingMy ZoneTopic filtersFind it atSidebar → Docs

Records & History

2 features

Every version and every decision, kept where the diagram lives.

17·Records & HistoryFree + Pro

Versioning & Publishing

A draft you can experiment in, and versions you can trust.

2
version kinds
2.5s
auto-save debounce
Snapshot
cost + health
1-click
switch back

Work freely in an auto-saved draft, then publish a labelled version that snapshots the topology, the cost and the health score together.

  • Draft holds your in-progress state; publishing freezes it as a numbered version with a changelog message
  • Every published version stores node and edge counts, estimated monthly cost, and the health result at that moment
  • Switch to any previous version, or discard the draft and snap back to what was last published
  • Version drawer lists the full history with who published it and when
  • AI-generated Markdown design doc exports from any published version
  • Free plans keep five versions per diagram; Pro keeps all of them
Draft / publishedChangelogVersion switchDiscard draftDesign doc exportFind it atCanvas → version drawer
§
18·Records & HistoryFree + Pro

Architecture Decision Records

Why the system looks like this, written down where it belongs.

4
ADR sections
Node
level anchoring
AI
drafting & search
Status
lifecycle

Structured ADRs anchored to a diagram — or to a single node on it — so the reasoning survives the people who made it.

  • Context, decision, alternatives considered, and consequences — the standard ADR shape, enforced
  • Attach an ADR to a diagram or pin it to a specific component on the canvas
  • Proposed / accepted / superseded status so a reversed decision stays legible
  • The AI can draft an ADR from the conversation you just had about a tradeoff
  • It can also search your existing ADRs, so it answers with your team's history instead of generic advice
  • ADRs live as notes, which means folders, Markdown export, and full-text editing come free
ContextDecisionAlternativesConsequencesAI draftAI searchFind it atNotes → ADR, or canvas node → ADR

Workspace & Account

2 features

The shell around the work: overview, auth, credits, and plans.

19·Workspace & AccountFree + Pro

Workspace Dashboard

Every diagram, its cost, and its score on one screen.

3
portfolio stats
Live
SVG thumbnails
Tags
and starring
Trash
soft delete

The landing surface: portfolio-level numbers up top, thumbnailed diagrams below, templates one click away.

  • Totals for diagram count, estimated monthly cost across everything, and average health score
  • Recent diagrams render as live thumbnails with their health colour and cost badge
  • Star what matters, tag what needs grouping, and filter down from there
  • Start-from-template row sits inline so a new diagram is two clicks away
  • Soft delete means a wrong click is recoverable, not fatal
  • Collapsible sidebar spanning all five editors, the library, and account settings
Portfolio costAvg healthThumbnailsTagsStarTrashFind it atSidebar → Architecture
20·Workspace & AccountFree + Pro

Accounts, Plans & Billing

Sign in with one click. See exactly what you're spending.

$0
Free forever
$20
Pro / month
$16
Pro annual / month
3
auth methods

GitHub or Google SSO, transparent credit metering, and a Free tier that stays genuinely usable without a card.

  • GitHub OAuth, Google OAuth, or email and password — connect additional providers to one account later
  • Free: 3 diagrams, 20 nodes and 30 edges each, 3 notes, 5 versions, 50 AI credits, PNG export
  • Pro: unlimited diagrams, nodes, notes and versions, 500 credits, Deep Review, scan-import, custom templates, share links, SVG / Mermaid / JSON export
  • Credit usage is itemised by action with call counts and the model each one ran on
  • Razorpay checkout, cancel any time, annual billing saves $48 a year
  • Light and dark themes across the whole product
GitHub SSOGoogle SSOEmailRazorpayLight / darkFind it atSidebar → Plans · Billing & Usage
In active development

What's coming next.

The canvas was built multiplayer-ready from day one. These are the pieces landing on top of it — listed separately from the catalog above, because they aren't shipped yet.

👥
Live presence cursors

See teammates' cursors, names and selections in real time — no more asking in Slack who's editing what.

Y.js CRDT engine

Concurrent edits merged mathematically instead of locked. No merge conflicts, no last-write-wins surprises.

💬
Inline comments

Threads anchored to a specific node or edge, resolved as decisions land, kept alongside the diagram.

🔐
Granular access control

View-only, comment-only or full edit per diagram, with org-level roles for who can create and share.

🧱
Terraform & Pulumi export

Turn a validated diagram into IaC scaffolding instead of retyping it into HCL by hand.

📡
Drift detection

Compare what's actually running against what the diagram claims, and flag where reality diverged.

Where we're going

Architecture as institutional memory.

Today BackArch is a design and validation tool. The goal is the single source of truth for every architectural decision your organisation has made — searchable, connected, and still legible three years later.

Shipping next
🔨 Building
  • Real-time collaboration on the canvas (Y.js CRDT)
  • Live presence cursors and selections
  • Inline comments anchored to nodes and edges
Planned
📋 Planned
  • Terraform and Pulumi export
  • GitHub integration and architecture diff view
  • Team workspaces with granular access control
On the horizon
🔭 Vision
  • Drift detection — live infrastructure against the diagram
  • AI cost forecasting from simulation runs
  • Compliance report generation
Ready to build better systems?

Your architecture deserves
the right tools.

Free forever for solo developers. No credit card, no setup friction — open the canvas and start designing the system you've been thinking about.

Free forever · 3 diagrams · SSO with GitHub or Google
Got a question, an idea, or hit a bug?
We're a small team and we read every message. Drop us a line at support@backarch.com and we'll get back to you, usually within a day.