Pricing

You pay for the model.
Nothing else.

Code Claw is free to download. There's no seat, no subscription to us, and no markup on tokens. Run free on your own machine, or bring a key and pay the provider directly.

Free forever

Run it locally

$0no key, no bill
  • Install an open coding model in one click
  • Runs entirely on your machine — private and offline
  • No token bill, no seat, no account

Best for privacy, side projects, and zero budget.

Bring your own key

Pay the provider direct

wholesaleno markup from us
  • Anthropic, Gemini, OpenAI, Kimi, DeepSeek, GLM
  • You pay the provider their list price — we add nothing
  • Switch model per task; a live counter shows spend

Best for controlling exactly what you spend.

Managed

We handle the key

one plannothing to set up
  • A cheap, capable model under the hood
  • No provider accounts or keys to manage
  • The fastest way to get started

Best for getting going with zero setup.

Made concrete

What a month actually costs

One developer, everyday use — edits, fixes, tests, and questions. Here's the range, depending on how you run it.

Local model (Qwen)$0
BYOK · DeepSeek~$5–10
BYOK · Gemini Flash~$5–12
BYOK · premium frontier~$40–120
Flat AI-IDE subscription$20

Illustrative only. Actual cost depends heavily on how much you use it and which model you pick; token prices are set by each provider and change over time.

Pick per task

A model for every job

Most work doesn't need a premium model. Code Claw lets you match the model to the task and switch any time — so you spend big only when it's worth it.

Frugal~$0.30–0.60
per 1M input tokens

Pennies per million tokens. More than enough for edits, refactors, tests, and everyday questions — where most of your work lives.

DeepSeek · Gemini Flash · GLM · Kimi

Balanced~$2–3
per 1M input tokens

A step up in reasoning for trickier features and gnarlier debugging, still a fraction of a premium model.

OpenAI GPT-4o · mid-tier frontier

Premium~$10–15
per 1M input tokens

Reach for it on the hardest problems — architecture, subtle bugs — then drop back down. You decide, per task.

Top frontier models

Approximate public list prices, for direction only. Output tokens are priced separately — check each provider for current rates.

No markup. No lock-in. Ever.

When you bring your own key, your tokens are billed by the provider at their price — we never sit in the middle. Change models, change providers, or go fully local whenever you like.

Start free, upgrade only if you want to

Download Code Claw and run your first task on a local model for $0.

Download Code Claw