SAASPOCALYPSEverdict #KIRO-8F2C

scanned 2026.04.30 · 14:51

subject of investigation

kiro.dev

Item: kiro.dev
Rating: 12
Author: saaspocalypse

▸ agentic AI IDE with spec-driven development

verdict: DON'T

buildability score

/100

tier · don't

the blunt take

“You're not building an AI IDE this weekend. You're not building one this decade. Cursor took a team of ex-OpenAI engineers years. Kiro is built on Code OSS, ships a CLI, runs autonomous agents, and integrates frontier models. The moat isn't the LLM call — it's the 200,000 lines of IDE plumbing you haven't written yet.”

The "interesting feature" here is a full VS Code fork with custom agent orchestration, spec-to-code pipelines, MCP integration, file-event hooks, and a headless CLI. That's not a feature. That's a company. Several, actually.

cost breakdown.

their price ←→ your price

what they charge●

Free tier (pricing page exists)

/ user/mo

※ Free tier visible; paid tiers implied by enterprise nav but not shown on homepage

annual:$0

what it costs you✦

01 · Claude Sonnet 4.5 / frontier LLM API (per-prompt, per-user, at scale)??? — scales with usage

02 · Code OSS fork infra (build pipeline, auto-update server, artifact hosting on S3/CloudFront)??? — depends on download volume

03 · Auth (GitHub, Google, AWS IAM Identity Center OAuth)$0.00

04 · Postgres (user accounts, credits, spec storage)$25.00

05 · Vercel / CloudFront (Next.js marketing site)$0.00

06 · Domain$1.00

07 · Your remaining sanitypriceless

TOTAL / mo$26.00 + usage

▸ break-even:approximately never — the LLM API bill alone will eat you alive before you ship v0.1

moat

how deep is the moat.

methodology →

6.0/10

aggregate score · meaningful

weighted average of the six axes below. higher = harder for an indie hacker to displace.

real moat

capital

7.0/10

what it costs to keep the lights on

technical

9.2/10

depth of the underlying engineering

network

0.0/10

users compound users

switching

8.0/10

stickiness of customer data + workflow

data

4.0/10

proprietary data accumulates over time

regulatory

0.0/10

real licenses + compliance, not SOC 2 theater

or, you know, use one of these.

if building feels spicy

option A

Cursor (just use it)

Already exists. Already great. Already has spec-ish features via rules. You are not beating this.

option B

Continue.dev (self-host)

Open source AI coding assistant. Plugs into VS Code or JetBrains. Bring your own model. Ships today.

option C

Aider (CLI, open source)

Terminal-native agentic coding. Supports every model. No IDE fork required. Actually shippable by one person.

what'll actually be hard.

est. total: ∞

▸ 6 months forking Code OSS · 6 months on agent orchestration · 6 months on spec pipeline · the rest of your life on latency

easy

medium

hard

nightmare

easy

Marketing site

Next.js + CloudFront. You already know this. Not the problem.

medium

LLM prompt pipeline for spec generation

Prompt → EARS requirements → architecture → task list. Doable in a weekend. The output quality is the hard part.

hard

Agent hooks system (file-event triggers)

File watcher → agent dispatch → background execution loop. Async state management gets gnarly fast.

hard

MCP integration (remote + local)

Model Context Protocol is young, docs are thin, and you need to handle auth, transport, and schema negotiation for arbitrary third-party servers.

nightmare

Forking and maintaining Code OSS

VS Code is 1.5M+ lines of TypeScript. Keeping your fork current with upstream while shipping custom agent UX is a full-time job for a team.

nightmare

Autonomous agent orchestration at IDE scale

Multi-step tool use, file diffs, rollback, approval flows, credit metering per prompt, autopilot mode — this is the entire product and it never ships clean.

detected signals· we measured these

frameworkNext.jscdnCloudFront

recommended stack · inferred

Code OSS fork (Electron + TypeScript)Claude Sonnet 4.5 / Anthropic API (??? cost)Next.js 15 + CloudFront (marketing site)Postgres (user accounts, credit ledger, spec storage)Custom CLI (Go or Rust, cross-platform)

ready to build?

We'll email you the MVP guide. It won't be the original. But it'll ship.

▸ generated with love, by a heartless robotverdict v2.1 · saaspocalypse.dev