Cheaper. Faster. Cleaner AI coding

Distill.codes helps Claude Code finish tasks with fewer output tokens, fewer operations, and cleaner core code.

Choose plan and start 3-day trial

Try it on your real coding tasks.
Compare cost, speed, and code size.

Cheaper

Lower output-token waste and provider spend.

Faster

Fewer unnecessary loops and tool calls.

Cleaner

Smaller patches with less review and maintenance.

Baseline Leaner generated changes

Code bloat compounds.

Extra edits and lines become review and maintenance debt. Distill.codes keeps generated changes leaner over time.

Your keys stay yoursUse your own provider credentials.

Private by designNo prompts or code stored.

Drop-in setupOne config line for Claude Code.

Simple pricing

Personal

$4/mo

For individuals

$40/year

Company

$40/mo

For teams

$400/year

Choose plan and start 3-day trial

Benchmarks

Internal Claude Code run. Verify the effect on your own repo with the CLI below.

65.5%

fewer output tokens

Fable 5 xhigh, verified task

63.6%

faster

Same model, same prompt

48.0%

less source LOC

Smoke-verified implementation

Run the paired CLI benchmark below with your own proxy URL.
Results vary by task, model, and agent behavior.

View benchmark results

Run a quick benchmark

Use one command to compare Claude Code direct vs Distill.codes in separate local folders.
Create an account and start the 3-day trial to get your proxy URL.

npx distill-codes bench "https://proxy.distill.codes/<proxy-key>/essential/anthropic"

Start 3-day trial and get your proxy URL

Savings calculator

Estimate monthly savings from a conservative output-token reduction.

Monthly AI coding spend

Distill.codes planPersonal $4/moCompany $40/mo

$100

Monthly spend

Current provider cost

$50.00

Provider savings

50% conservative output-token reduction estimate

$46.00

Net savings

After Personal plan ($4.00/mo)

Calculator uses a conservative estimate, not a guarantee.
Actual savings vary by task, model, provider pricing, and agent behavior.

Measure it on real tasks

Run the same task with and without Distill.codes in separate fresh directories.
Compare output tokens, wall-clock time, changed LOC, and final result quality.

Use the CLI benchmark above for a quick paired run.

How it works

Distill.codes does three simple things:

1Sits between your coding agent and your LLM provider.
2Lets you connect Claude Code to the Distill.codes proxy using your existing provider credentials or agent login.
3 Applies Essential Mode to supported generation requests.

Requests continue to your provider.

Read Documentation

AI coding guides

Set Up Claude Code with Distill.codes AI Coding Agent Cost Optimization Reduce Claude Code Token Usage and Cost AI-Generated Code Bloat What Is an AI Coding Proxy?

View all guides

Built by an optimizer

Distill.codes is built by Viacheslav Bogdanov, founder of MyDream Labs and a software engineer with 25+ years of experience shipping, refactoring, and optimizing real products.

About the company

Planet-friendly by design

Leaner agent runs use less compute.
Less compute means less energy used, so Distill.codes helps you build with a smaller footprint.