Cognition AI’s AI-native code editor. A VS Code fork with inline autocomplete, a command bar, and Cascade: a multi-file agentic editor with full codebase context.
Owned by Cognition since July 2025, after Google took a licensing deal and an earlier OpenAI acquisition collapsed. Now a sibling product to Devin inside the Cognition stack.
May 13, 2026: Cognition shipped SWE-1.6 as Windsurf’s latest default coding model, and the pricing page now lists a Light plan with unlimited usage alongside Free, Pro, Max, and Teams. Source: Windsurf pricing (reverified 2026-05-13).
May 6, 2026: ServiceNow Build Agent reached Windsurf and other major AI coding tools, bringing ServiceNow platform context into agentic coding workflows while keeping ServiceNow governance and deployment controls in scope.
May 1, 2026: The MCP STDIO command-execution disclosure named Windsurf in the highest-risk IDE bucket, with a reported zero-click prompt-injection path to local RCE. Treat MCP auto-registration and third-party servers as privileged shell access until patched and sandboxed.
April 17, 2026: Windsurf shipped Claude Opus 4.7 support within 24 hours of Anthropic’s release, alongside Cursor, Zed, and Continue.
System Verdict
Pick Windsurf for agentic multi-file editing with real model choice and Cognition-stack integration. Cascade reads the repo, plans cross-file edits, runs terminal commands, and ships with access to OpenAI frontier models, Claude Opus 4.7, Gemini 3.1 Pro, and Cognition’s own SWE-1.6.
SWE-1.6 is Cognition’s latest proprietary coding model, succeeding SWE-1.5 as the default. It targets the same fast iterative-editing workload while extending agentic depth on multi-file refactors.
Skip it if Cursor already fits. Cursor has more polish, a larger extension community, and comparable agent depth. Windsurf’s edge is the Cognition ecosystem: tickets delegated to Devin flow back into the IDE cleanly.
Who pays which tier: Free for evaluation, Pro $20/mo for most individuals, Max $200/mo for sustained heavy use, Teams $40/user for shared admin, Enterprise for SSO and hybrid deployment.
Key Facts
| Vendor | Cognition AI (acquired July 2025) |
| Product class | VS Code fork, AI-native IDE |
| Flagship agent | Cascade, multi-file editor with repo awareness |
| Default model | SWE-1.6, Cognition proprietary, latest as of May 2026 |
| Other models | OpenAI frontier models · Claude Opus 4.7 · Gemini 3.1 Pro · local via Ollama |
| Pro pricing | $20/mo, unlimited at API price |
| Max pricing | $200/mo, priority support, unlimited at API price |
| Light plan | Unlimited usage tier listed on the pricing page (verify entitlements before standardizing) |
| Teams pricing | $40/user/mo with admin dashboard |
| Enterprise | Custom with SSO, RBAC, hybrid deployment |
| Free tier | Yes, with daily and weekly refresh quotas |
Every data point above was verified against vendor documentation on 2026-05-13. See Sources.
What it actually is
A VS Code fork with AI baked into the editor shell. Tab autocomplete handles inline suggestions. A command bar (Cmd+I) takes natural-language edit requests. Cascade handles multi-file agentic edits with full codebase context.
Model selection is global or per-session. SWE-1.6 is the latest Cognition default, replacing SWE-1.5 for new sessions. Heavy reasoning work routes to Claude Opus 4.7 or OpenAI frontier models. Gemini 3.1 Pro is available for long-context runs. Ollama covers local models.
The Cognition acquisition closed the loop with Devin. Tickets delegated to Devin can round-trip through Windsurf. The IDE also consumes the MCP Marketplace from the Cognition platform.
VS Code extension compatibility holds. Keybindings, themes, and most extensions transfer from a standard VS Code install.
When to pick Windsurf
- You want agentic multi-file edits with real model choice. Cascade plans and executes across files. SWE-1.6 is fast; Opus 4.7 and OpenAI frontier models cover the heavy work.
- You already use or plan to use Devin. Cognition’s Devin-to-Windsurf handoff is tighter than any competing IDE’s third-party agent integration.
- You want VS Code compatibility with deeper AI than Copilot. Extensions and keybindings port cleanly. Cascade outruns Copilot’s agent mode on multi-file refactors.
- You value SWE-1.6’s speed on iterative edits. The proprietary model prioritizes latency for fast-turn refactors.
- You are cost-conscious. Free tier with daily and weekly refresh quotas is generous enough to test realistic workflows.
When to pick something else
- Maximum IDE polish and community: Cursor. More extensions, larger community, comparable agent depth.
- Terminal-first agentic coding: Claude Code. Strongest CLI agent, flat $100/mo on Max 5x.
- Open-source BYOK inside VS Code: Cline with Plan/Act modes. No subscription floor.
- Inline autocomplete only, cheapest option: GitHub Copilot at $10/mo.
- Async ticket delegation: Devin (same owner). Hand work off, review the PR.
Pricing
Pricing via windsurf.com/pricing. Paid tiers move to unlimited usage billed at API price with tier-specific features and support.
| Plan | Price | Usage | Features | Who’s it for |
|---|---|---|---|---|
| Free | $0 | Daily/weekly refresh | Basic Cascade, Tab autocomplete | Evaluation |
| Light | Listed | Unlimited usage | Newer tier; verify exact entitlements on pricing page | Light agentic users |
| Pro | $20/mo | Unlimited at API price | Premium models · Fast Context · SWE-1.6 | Most individuals |
| Max | $200/mo | Unlimited at API price | All Pro features plus priority support | Sustained heavy workloads |
| Teams | $40/user/mo | Unlimited at API price | Admin dashboard · analytics | Small teams |
| Enterprise | Custom | Unlimited | SSO · RBAC · hybrid deployment | Compliance-heavy orgs |
Prices verified 2026-05-13 via Windsurf pricing. SWE-1.6 is Cognition’s latest default coding model.
Against the alternatives
| Windsurf Pro | Cursor Pro | Cline (open source) | |
|---|---|---|---|
| Price | $20/mo | $20/mo | Free (BYOK API costs) |
| Form factor | VS Code fork | VS Code fork | VS Code extension |
| Flagship agent | Cascade | Composer / Agent | Plan/Act modes |
| Model choice | SWE-1.6 · OpenAI frontier models · Opus 4.7 · Gemini | Mostly bundled | Any BYOK provider |
| Extension ecosystem | VS Code compatible | VS Code compatible | Native VS Code |
| Best viewed as | Cognition-stack IDE | Polished default IDE | Free agentic extension |
Failure modes
- UI lags behind Cursor on polish. Occasional 3-5 second agent pauses on context-heavy refactors. Cursor often handles the same work in under 2 seconds.
- Smaller third-party ecosystem than Cursor. Niche extensions sometimes ship for Cursor first.
- Cognition dependency. Windsurf’s roadmap now tracks Cognition’s priorities. Devin-IDE integration is a feature, vendor-lock is the trade-off.
- SWE-1.6 quality is workload-dependent. Fast on iterative edits. Frontier models still beat it on novel reasoning-heavy tasks; published 1.6-vs-1.5 deltas are still vendor-reported.
- Free tier quotas are daily and weekly. Heavy evaluation hits the ceiling quickly. Budget a Pro month to run a real benchmark.
- MCP zero-click risk is a watch item. The May 1 MCP STDIO report says Windsurf had a reported zero-click path from prompt injection to local RCE. Disable auto-registration, review active MCP configs, and sandbox servers.
- Teams pricing is not bundled with Devin. The two products bill separately. Multi-product teams should model total cost before switching.
- Post-acquisition leadership churn. Founding leads moved to Google in 2025. Release cadence is stable so far but remains a watch item.
- Model pricing variance. “Unlimited at API price” means heavy users on Opus 4.7 will see real API costs land on their Cognition invoice.
Methodology
This page was produced by the aipedia.wiki editorial pipeline, an automated system that ingests vendor documentation, verifies pricing and model details against primary sources, and generates the editorial analysis you are reading. No individual human wrote this review. Scoring follows the four-dimension rubric at /about/scoring/ (Utility × Value × Moat × Longevity, unweighted average). Last verified 2026-05-13 against Windsurf pricing, Windsurf model docs, and the Cognition acquisition announcement.
FAQ
Who owns Windsurf? Cognition AI acquired Windsurf in July 2025 after Google secured a licensing deal for Windsurf’s technology and an earlier OpenAI acquisition attempt collapsed over Microsoft IP-rights disputes. Cognition owns the IP, product, brand, and team.
What is SWE-1.6? Cognition’s proprietary coding model, succeeding SWE-1.5 as the default in May 2026. Optimized for speed on iterative edits; the predecessor SWE-1.5 was claimed roughly 13× faster than Sonnet 4.5 at near Claude 4.5-level coding-benchmark performance. Published 1.6-vs-1.5 deltas remain vendor-reported.
Which external models does Windsurf support? OpenAI frontier models (OpenAI), Claude Opus 4.7 (Anthropic), Gemini 3.1 Pro (Google), plus local models via Ollama. Selectable per session or globally. See Windsurf model docs.
How does Windsurf compare to Cursor? Same $20/mo Pro price. Cursor has more polish and a larger community; Windsurf has tighter Cognition-stack integration (Devin hand-off) and broader model choice inside Cascade. Most individuals see comparable agent depth on multi-file work.
Can I keep my VS Code extensions? Yes. Windsurf retains VS Code extension compatibility, keybindings, and settings sync. Most extensions install without modification.
Sources
- Windsurf pricing: current plans and usage model
- Windsurf model docs: supported models and configuration
- Cognition: acquisition of Windsurf: ownership and acquisition details
- CNBC: Cognition valued at $10.2B post-Windsurf: post-acquisition valuation
Related
- Category: AI Coding
- Comparisons: Cursor vs Windsurf
