Windsurf

Cut Your Windsurf Token Costs by 40–70%

Windsurf's Cascade agent bundles your files, codebase, and terminal into every request. Terse compresses the one thing you control — the instruction you type into Cascade — and tracks the token cost of each request. All on-device.

Start 30-Day Free Trial See How It Works
40–70%instruction compression
AutoWindsurf detection
Any modelper-request cost tracking
0data leaves device

How Terse Works with Windsurf

Windsurf is an AI IDE, and its Cascade agent is flow-aware — it reads your open files, the wider codebase, and the terminal, then edits across multiple files. Most of that context is added automatically. Terse compresses the part you type into the Cascade input and shows what each request costs, so you can spot the turns that burn credits fastest.

Windsurf — Cascade
Terse Active
Cascade · model: Claude Sonnet 4.5 · flow: 3 files open, repo context on
● Terse detected Windsurf — attached to Cascade input (Sonnet 4.5 · $3.00/M input)
You › could you please go ahead and refactor the authetication middleware so that it hopefully uses asynchrnous handlers, and also make sure to keep it backwards compatable with the existing api consumers if thats possible
Terse: 46 tokens → 19 tokens · "refactor auth middleware to async handlers, keep backwards compatible with existing API consumers" · 59% reduction
Cascade flow: reading 3 open files + retrieved context [~14,200 tokens]
⚠ Terse: this Cascade request is context-heavy (~14K tokens) — a leaner instruction won't shrink the flow, but it keeps your turn tight
14.2Krequest tokens
14.5Kwithout Terse
59%instruction cut
$0.04this request
⚡

Auto Windsurf Detection

Windsurf is built on VS Code, so Terse recognizes it the same way it handles Cursor and Copilot. Focus the Cascade panel and a floating Terse bar appears automatically — no setup, no extension to install.

Automatic
🗜️

Cascade Instruction Compression

Every instruction you type into the Cascade chat runs through Terse's 7-stage pipeline: typo correction, filler removal, hedge stripping, and phrase shortening. Code blocks, file paths, and symbols pass through untouched.

7-Stage Pipeline
🌊

Flow-Aware Cost Signals

Cascade quietly pulls in your open files, retrieved codebase context, and terminal output. Terse flags when a request is context-heavy so you can see which flow actions are the ones eating your credits.

Flow Insight
💰

Model-Aware Cost Tracking

Terse reads the model selected in Cascade and estimates per-request cost at the correct rate — Sonnet, Opus, GPT-4o, Gemini 2.5, or SWE-1 / Cascade Base. See which prompts translate to the most credits.

Per-Request Visibility
📊

Session & Credit Summary

Across a Windsurf session, Terse totals your instruction tokens, estimated cost, and how many tokens compression saved — so you know what your Cascade flow actually cost by the end of the day.

Session Analytics
🔒

100% On-Device

Terse reads only the Cascade input via macOS Accessibility APIs and compresses locally in Rust. It never reads your source files, your codebase index, or Cascade's API traffic. No prompt text or code leaves your machine.

Zero Cloud

Supported Windsurf Models

Terse recognizes the model you select in Cascade and applies the correct per-million rate to every request.

Claude Sonnet 4.5
$3/M input · $15/M output
Claude Opus 4
$15/M input · $75/M output
GPT-4o
$2.50/M input · $10/M output
Gemini 2.5 Pro
$1.25/M input · $10/M output
GPT-4o mini
$0.15/M input · $0.60/M output
Gemini 2.5 Flash
$0.075/M input · $0.30/M output
Claude Haiku 4.5
$0.80/M input · $4/M output
SWE-1 / Cascade Base
Windsurf in-house · credit-based usage

Frequently Asked Questions

Everything about using Terse with Windsurf.

Does Terse work with Windsurf?
Yes. Windsurf is a VS Code–based AI IDE, so Terse attaches to the Cascade chat input the same way it does for Cursor and Copilot. Focus Windsurf and a floating Terse bar appears, compresses the instruction you type into Cascade, and tracks the per-request token count. Detection is automatic — no configuration.
Why does Cascade use so many credits/tokens?
Cascade is flow-aware: every request bundles your open files, retrieved codebase context, conversation history, and terminal output into the prompt. A single flow action can be tens of thousands of tokens, and Windsurf bills credits against that. Terse can't shrink the auto-added context, but it compresses your typed instruction 40–70% and surfaces the per-request cost.
Does Terse read my Windsurf code?
No. Terse only reads the text you type into the Cascade input, via macOS Accessibility APIs, and compresses it on-device. It does not read your source files, Windsurf's codebase index, or Cascade's internal API traffic. Nothing you type or open ever leaves your machine.
Which models does it track?
Whatever you pick in Cascade — Claude Sonnet 4.5, Opus 4, GPT-4o, Gemini 2.5 Pro, Claude Haiku 4.5, or Windsurf's own SWE-1 / Cascade Base. Terse applies the correct per-million rate to each request so the estimated cost matches your real usage.

Also Works With

Terse reduces token costs across every major AI coding tool.

⚡ Claude Code — CLI Agent → ⌨️ Cursor — AI Code Editor → 🤖 GitHub Copilot — VS Code → 🖥️ Aider — Terminal Agent →

Stop Burning Credits on Verbose Prompts

30-day free trial. No credit card required until your trial ends. Compresses your Cascade instructions across Claude, GPT-4o, Gemini, and SWE-1.

Start Free Trial — macOS Calculate Your Savings