Windsurf — Quick reference
AI-native code editor (built on VS Code) whose Cascade agent codes whole tasks across your project. Now owned by Cognition; windsurf.com redirects to Devin Desktop.
At a glance
What it is: a VS Code-based IDE whose centrepiece, Cascade, is an autonomous agent — it reads your whole project, plans multi-file changes, runs terminal commands, and iterates until the code works.
Where it runs: desktop app, download from windsurf.com (redirects to devin.ai/desktop).
Pricing: Free $0/mo unlimited Tab + inline edits, light agent quota · Pro $20/mo frontier models · Max $200/mo.
Setup & first edit
Install
Download Windsurf from windsurf.com and sign in. Built on VS Code — if you have used VS Code, the layout is familiar. The Free plan needs no credit card.
First Cascade task
Open a folder with your file, open the Cascade panel, and type a task in plain language. Cascade plans, writes the code, explains its logic, and runs it. Review the proposed edits before you accept.
I have a BLAST output file called blast_results.txt. Write a Python
script that reads it, extracts the top hit per query (e-value,
accession, description), and saves a clean CSV.
Key commands & shortcuts
| Action | What it does |
|---|
Tab | Accept the next-line autocomplete suggestion. Unlimited on every plan — never spends agent quota. |
| Inline edit | Select a block and describe a tweak right where the code is — no full Cascade task. Unlimited and free. |
Open Cascade ⌘L / Ctrl+L | Open the Cascade panel to start a task. |
| Cascade panel | Hand the agent a whole task; it reads your whole project, edits across files, and runs commands. |
Explain and Fix ⇧⌘. | Select an error in the editor — Cascade explains what broke and proposes a fix in place. |
| Send to Cascade | From the Problems panel, push the listed issues straight into the agent conversation for fixes in context. |
| Open diff | On any agent edit, see exactly which lines were added or removed before accepting. |
| Checkpoint revert | Roll the project back to a one-click snapshot taken at each Cascade step. |
The keyboard shortcuts are ⌘L / Ctrl+L (open Cascade) and ⇧⌘. (Explain and Fix, shown on the in-editor button); the other rows are commands/panels rather than fixed shortcuts.
Prompting patterns
Describe the outcome
Name a goal, not a function: "add a CSV export button", "refactor this analysis to use pandas". Cascade takes the whole task.
Be specific & located
Say what and where — "add a CLI argument for the input filename", "filter hits above 1e-5" — not "make it better".
Iterate, don't restart
Send follow-ups in the same conversation so Cascade keeps building on the code it already wrote.
Batch related tweaks
Group a few related changes into one message; each Cascade task draws on quota, so make them count.
Restate, don't pile on
If a request was too vague, restate it more precisely rather than adding more on top.
Reserve the agent
Use Tab + inline edits for small local stuff (rename, tweak a loop, add a docstring); save Cascade for multi-file work.
Context & modes
Whole-project context
Cascade reads your whole project, not just the open file, and stays aware of what you type and what runs in the terminal — so it can pick up mid-task.
Three ways to edit
Tab completions (predictive) → inline edits (selected block) → Cascade (whole, multi-file tasks). Pick the smallest tool that does the job.
Auto-fix
With Auto-fix on, when an edit introduces lint errors Cascade flags them (e.g. "5 new lint errors") and clears them itself before finishing.
Checkpoints
Each Cascade step leaves a snapshot. If a change goes wrong, revert to an earlier checkpoint in one click — nothing is silently overwritten.
Built-in integrations connect Cascade to GitHub, databases, Slack, and Figma. Pro and up add cloud agents, managed from the Agent Command Center Kanban board.
Pricing & limits
| Plan | Price | What you get |
|---|
| Free | $0/mo | Unlimited Tab completions + inline edits, light Cascade quota, limited model availability. Runs dry in ~2–3 real coding days of heavy agent use. No credit card. |
| Pro | $20/mo | Higher quotas; full model availability — OpenAI, Claude, Gemini frontier models, free Windsurf SWE 1.6 + open-source models, cloud agents. |
| Max | $200/mo | Significantly higher quotas for heavy users. |
| Teams | $80/mo + $40/seat | Collaboration, admin dashboard, priority support. |
| Enterprise | custom | SSO, centralised controls, deployment options. |
Tab completions and inline edits are unlimited and free on every plan; every Cascade task (including Send-to-Cascade, Explain-and-Fix, and Auto-fix) draws on the agent quota. Annual billing saves ~17–20%. Verify current numbers at devin.ai/pricing.
Gotchas & tips
• Free quota burns fast — heavy agentic sessions exhaust it in 2–3 days; upgrade to Pro only when the quota, not the editor, is the bottleneck.
• You still need coding basics — non-coders cannot meaningfully drive Cascade; you stay the reviewer who accepts edits.
• Always review before accepting — read the diff; use checkpoints as your safety net to revert and re-prompt sharper.
• Can spike CPU and slow down during long, complex agent runs.
• Product is in flux — post-acquisition rebrand to Devin Desktop means windsurf.com redirects to devin.ai; direction may shift.