Make the architecture call you can defend.

Describe a technology decision in plain English. In a few minutes you get a weighted trade-off matrix, the tensions that actually matter, and a recommendation written as an ADR you can paste straight into your repo.

Kafka

SNS+SQS

Outbox

Data integrityhigh

Ops simplicityhigh

Reliabilityhigh

Evolvabilitymedium

Costmedium

From vague debate to a documented decision

Describe the decision

Name the options and add the context you'd give a senior colleague: team size, stack, scale, time horizon, how hard it is to reverse.

Answer a few questions

Only if something material is missing, you get up to four sharp framing questions, once. Anything you skip becomes a stated assumption.

Get a defensible answer

A scored, weighted comparison and a clear recommendation, usually in 2–5 minutes. Stress-test the weights, add new facts as they arrive, and share it as a branded PDF.

Every analysis includes

A consistent method, not a chat transcript. Weights are set before options are scored, so the answer isn't reverse-engineered from a favorite.

◎

Decision frame

What's being decided, the real constraints, and what's out of scope.

⚖

Weighted characteristics

The quality attributes that matter for your context, ranked by importance.

▦

Scored matrix

Each option scored 1–5 per characteristic, with weighted totals you can challenge.

⇄

Tensions that matter

Where the options genuinely pull against each other, and what you give up.

⚑

Risk & reversibility

One-way or two-way door, so you know how much rigor the call deserves.

✎

ADR-ready summary

A concise decision record with explicit assumptions, ready to commit.

⇅

What-if weights

Re-weight any characteristic and watch the totals move. See exactly which changes would flip the answer, and how robust it is.

↻

Follow-ups as facts change

Add new pricing, a constraint or vendor commitments and it re-scores what changed. Promises are weighed as promises, not delivered features.

⎙

Branded PDF

A clean, shareable report for stakeholders and review boards, alongside a Markdown copy for your repo.

Simple, pay-per-analysis pricing

No subscription. Credits never expire. Every analysis includes a follow-up. If an analysis fails or stops early, the credit comes back automatically.

Free

2 full analyses

Start free

10 analyses

$0.90 per analysis

Get started

25 analyses

$19

$0.76 per analysis

Get started

Prices include applicable taxes. Secure checkout by Stripe.

Questions

What kinds of decisions work best?

Choices between two to five concrete technical options with real constraints: datastores, messaging, service boundaries, hosting, build vs buy, frameworks, auth, data pipelines. The more context you give (team, scale, horizon), the sharper the result.

How is this different from asking a chatbot?

It follows the same structured method every time: frame the decision, weight what matters, score each option, surface tensions and risk, then recommend. You get a consistent, comparable artifact rather than a conversation that drifts with how you phrase the question.

Can I add new information later?

Yes. On a finished analysis you can follow up with new pricing, a constraint, vendor commitments or another option, and it re-scores what changed and shows whether the recommendation moves. You can also ask questions about the result. Your first follow-up on each analysis is free; after that each one uses 1 credit.

How long does an analysis take?

Usually 2–5 minutes. You can leave the page; finished analyses stay in your history.

What happens to what I submit?

Your text is processed by Anthropic (the AI model) and stored on Cloudflare so you can see your history. It's not used for advertising or sold. Don't submit confidential employer or client information. Details are in the privacy policy.

Is the recommendation always right?

No. It's structured decision support, not a substitute for your judgment. The value is in making weights, scores and assumptions explicit so your team can challenge them.

What if an analysis fails?

Its credit is returned automatically. If a result is clearly broken, email tanmay.avinash.deshpande@gmail.com and you'll get the credit back. See the refund policy.

Loading…

What are you deciding?

A few questions first

Weighted scores

What would change the answer

Assumptions

Follow upFollow-ups

Full analysis