███████╗██╗    ██╗███████╗███████╗████████╗   ██╗
██╔════╝██║    ██║██╔════╝██╔════╝╚══██╔══╝   ██║
███████╗██║ █╗ ██║█████╗  █████╗     ██║      ██║
╚════██║██║███╗██║██╔══╝  ██╔══╝     ██║      ╚═╝
███████║╚███╔███╔╝███████╗███████╗   ██║      ██╗
╚══════╝ ╚══╝╚══╝ ╚══════╝╚══════╝   ╚═╝      ╚═╝

2x cheaper than Claude Code · Pay per use · No subscription

The Claude Code alternative.
Same tasks. Half the cost.

Sweet runs in your terminal just like Claude Code — reads your codebase, edits files, runs commands, writes tests. Pay only for what you use, no subscription required.

$ npm i -g @sweet-cli/sweet

Start for $5 Read the Docs

cheaper per token than Claude Code (Sonnet)

$0.86

to build a full auth system with tests

$0/mo

subscription — pay only for what you use

How it works

Same workflow as Claude Code.
Pay-per-use. No subscription.

Sweet works directly in your terminal — same tasks, same quality bar. You pay per token with no monthly commitment. We use cost-efficient models and pass those savings to you.

Step 01

Describe the outcome

Tell Sweet what you want to ship — a feature, a refactor, a bug fix, a migration. Plain English is fine.

Step 02

Sweet works autonomously

It reads your repo, makes a plan, edits files, runs commands, and iterates until the task is done. No hand-holding needed.

Step 03

Review what was built

Come back to a diff, passing tests, and a clear summary of what changed and why. Merge or ask for changes.

Real session

Watch it work.

Sweet reasons through the problem, uses tools to explore your codebase, writes the code, and runs the tests — all in one pass.

sweet — building auth system

› build a JWT auth system with login, register, and protected routes

read_file package.json, src/routes/index.ts, prisma/schema.prisma read_file src/middleware/*, src/models/user.ts
💭 Reasoning: The app uses Express + Prisma. No auth middleware exists yet. I'll add bcrypt + jsonwebtoken, create the User model migration, build the auth routes, and add a protect() middleware. Tests with Jest + supertest.
write_file src/auth/jwt.ts — JWT sign/verify helpers write_file src/auth/middleware.ts — protect() route guard write_file src/routes/auth.ts — POST /login, /register run_command npx prisma migrate dev --name add_users write_file tests/auth.test.ts — 12 test cases run_command npm test
✓ 12 tests passing
✓ 6 files written → auth/jwt.ts, auth/middleware.ts, routes/auth.ts, migrations/…
time: 4m 38s · tokens: 218k · cost: $0.93

What you can do

Works on real engineering tasks.

🔧

Build features end to end

Give Sweet a goal. It reads the existing codebase, writes the implementation, updates tests, and wires everything together.

$ sweet start "add Stripe subscription billing with webhook handling and invoice history page"

🐛

Fix bugs with context

Point Sweet at a failing test or error log. It traces the issue through your codebase and writes the fix with a regression test.

$ sweet start "users are getting logged out randomly — investigate the session handling"

🧹

Refactor safely

Break up a god class, migrate a database schema, or modernize a legacy module — with checkpoints and tests to verify correctness.

$ sweet start "refactor the payment module from callbacks to async/await, keep all tests green"

⚡

Run autopilot for hours

Set Sweet to work for a defined window. It plans a queue of tasks, executes them in sequence, and surfaces a summary when time's up.

$ sweet workfor 2h "clear the backlog: fix linting errors, update deps, add missing tests"

Cost comparison

Same tasks. Different price.

Open source models via Fireworks AI. You see every token and every dollar.

Task	Sweet	Claude Code	Cursor
Auth system with tests (~200K tok)	$0.86	$1.34	$1.34 + sub
REST API + validation (~150K tok)	$0.64	$1.01	$1.01 + sub
React dashboard (~250K tok)	$1.07	$1.68	$1.68 + sub
Bug fix + tests (~80K tok)	$0.34	$0.54	$0.54 + sub
Monthly (daily use)	$25–50	$20–200/mo sub	$20/mo + usage

Sweet uses GLM 5.1 via Fireworks AI. Claude Code costs based on Sonnet 4.6 API pricing. Token estimates are approximate for typical agentic coding tasks.

Pricing

Pay for what you use.

No subscriptions. Prepaid balance, pay only for what you use. Top up anytime.

Try it out

Good for 10–15 tasks. Enough to know if Sweet fits your workflow.

Start with $5

Start building for $5.

No subscription. No lock-in. Open source models. Transparent pricing. Your first task runs in under 2 minutes.

Get Started Read the Docs

The Claude Code alternative.Same tasks. Half the cost.