What's happening to Cascade on July 1, 2026?

Cascade — the original local AI agent that powered Windsurf — reaches end-of-life on July 1, 2026, exactly 14 days from today (June 17, 2026). Cognition rebranded Windsurf as Devin Desktop on June 2, 2026 and is replacing Cascade with Devin Local, a Rust rewrite that's roughly 30% more token-efficient and supports sub-agents. After July 1, Cascade stops receiving updates and Cognition will encourage migration to Devin Local. Your existing Devin Desktop installation will continue working, but Cascade-specific features, prompts tuned for Cascade's quirks, and any Cascade-only integrations will break or degrade. The forced migration is a small but real disruption for the meaningful number of teams who built workflows around Cascade specifically.

Devin Local vs Cursor vs Claude Code — which should I migrate to?

Devin Local is the lowest-friction choice if your team is already on Devin Desktop. It's a Rust rewrite of Cascade with the same surface area, 30% better token efficiency, sub-agent support, and integration with the Agent Command Center. Cursor is the right migration target if you valued Windsurf's IDE-first surface but want a more programmable agent platform — the June 2026 Cursor SDK release adds custom tools and stores, making it a true agent IDE. Claude Code is the right migration target if your team values terminal-first workflows and tight Claude-model integration. Honest recommendation: stay on Devin Desktop with Devin Local if you're happy with the IDE; switch to Cursor if you want more programmability; switch to Claude Code if you want to escape the IDE and live in the terminal. None of the three are wrong choices.

Will my Cascade configuration and rules carry over to Devin Local?

Mostly yes, with caveats. Your .windsurfrules file (now treated as .devinrules or equivalent) carries over. Cascade workflow files in the workspace carry over. The Agent Command Center in Devin Desktop reads your existing configuration and migrates it on first launch. The caveats: (1) prompts that exploited specific Cascade-engine quirks may behave differently — Devin Local is a different model architecture, so some prompt patterns will need re-tuning, (2) Cascade-specific tool definitions need to be updated for the Agent Client Protocol (ACP) that Devin Local uses, (3) any third-party integrations that hooked into Cascade's internal API will need updating. Plan for 1-3 days of light re-tuning for a typical small-team migration. Larger teams with heavy customization should plan for a week.

Should I migrate before July 1 or after?

Before. Cognition has been pushing migration since June 2 and Devin Local has been stable in production for two weeks. Migrating before the EOL means you can roll back to Cascade if you hit issues — after July 1, that fallback disappears. Realistic migration window: any time in the next 7-10 days. Use the rest of June 17-30 to test Devin Local on a non-critical workflow, validate that your rules and tools carry over, then flip the default before the holiday week starting June 28. Avoid migrating during the first week of July when Cognition will be flooded with last-minute migrations and any support issues will queue up.

What's the broader story here — is Cognition consolidating the AI coding tools market?

Yes. The Windsurf-to-Devin-Desktop rebrand was Cognition signaling that the IDE is now just the supervised surface for the Devin agent family, with Devin Cloud as the autonomous-task counterpart. The same week as the Cascade EOL (June 2026), Cursor is reportedly being acquired by SpaceX in a $60B all-stock deal — another major consolidation. Anthropic shipped Claude Code as a first-class terminal IDE rivaling Cursor. OpenAI shipped Codex as a coding-specific surface. The market is collapsing from 'AI features bolted onto IDEs' into 'IDEs purpose-built for agentic coding,' with Devin Desktop, Cursor, and Claude Code as the three main contenders. The Cascade EOL is one small marker in a much larger 2026 consolidation.

Quick Answer

Cascade EOL July 1, 2026: Migrate to Devin Local, Cursor, or Claude Code?

Published: June 17, 2026

Cascade EOL July 1, 2026: Migrate to Devin Local, Cursor, or Claude Code?

Cascade — the original Windsurf local agent — reaches end-of-life in 14 days. Cognition is replacing it with Devin Local inside the rebranded Devin Desktop. If you’re a Windsurf or Devin Desktop user, you have a forced migration this month. If you’ve been watching from a different IDE, this is a natural moment to evaluate whether to switch. Here’s a practical comparison of the three migration targets.

Last verified: June 17, 2026.

TL;DR

Cascade EOL: July 1, 2026 (14 days from today)
Default migration path: Cascade → Devin Local (same IDE, Rust rewrite, 30% more efficient)
Alternative 1: Cursor — more programmable, June 2026 SDK adds custom tools
Alternative 2: Claude Code — terminal-first, tight Claude integration
Recommendation: Stay on Devin Desktop unless you have a specific reason to switch
Migrate before: June 25 to leave time for testing before EOL week

What’s actually changing

Cognition acquired Windsurf in July 2025. On June 2, 2026 they rebranded the IDE to Devin Desktop and announced the Cascade-to-Devin-Local migration. The reframing:

Devin Desktop = the supervised IDE surface (formerly Windsurf)
Devin Cloud = the autonomous remote agent (the original Devin product)
Devin Local = the local-execution agent inside Devin Desktop (replacing Cascade)
Agent Command Center = the new default surface in Devin Desktop, managing both local and cloud agents

The Rust rewrite of Cascade to Devin Local is the substantive technical change. The 30% token efficiency improvement is real and measurable. Sub-agent support is new. The Agent Client Protocol (ACP) — open standard for editor-agent integration — means you can plug in Codex, Claude Agent, or other ACP-compatible agents alongside Devin Local.

The three migration targets

Devin Local (stay on Devin Desktop)

Best for: Teams already happy with Windsurf / Devin Desktop. Lowest friction.

What you get:

30% better token efficiency vs Cascade
Sub-agent support
ACP compatibility (plug in other agents)
Same IDE surface, settings migrated automatically
Agent Command Center as the default surface

What you give up:

Cascade-specific prompts and tool definitions need light re-tuning
Locked into Cognition’s roadmap for the IDE

Effort to migrate: 1-3 days for typical teams. Most config carries over.

Cursor (switch IDEs)

Best for: Teams that want a more programmable agent platform, or that find Cognition’s roadmap limiting.

What you get:

June 2026 Cursor SDK with custom stores and custom tools (TypeScript / Python functions exposed to the agent)
Bugbot for automated code review — faster and cheaper as of June 2026
Improved Design Mode (Cursor 3.7) with multi-select and voice input
Broad model selection (Claude, GPT, Gemini, plus open-weight via API)
Reportedly training its own frontier model (announced May 2026)

What you give up:

Lose the Cognition-specific autonomous Devin Cloud integration
Higher subscription cost than Devin Desktop in most configurations
Forthcoming SpaceX acquisition adds uncertainty about future direction

Effort to migrate: 3-7 days for typical teams. Different IDE surface, different rules format, but mostly familiar AI-coding patterns.

Claude Code (switch to terminal-first)

Best for: Terminal-first developers, teams with heavy Anthropic investment, agentic-workflow purists.

What you get:

Terminal-native TUI — no IDE required
Tight Claude Opus 4.7 / Sonnet 4.5 integration (Fable 5 was the headline model until the June 12 export-control suspension)
Background sub-agents, skills, dynamic workflows
The cleanest interface for headless / CI usage via claude -p
Free with paid Claude plans (Pro / Max) for interactive use

What you give up:

No IDE surface — you bring your own editor (most users keep VS Code or Neovim open alongside)
Tightly Anthropic-locked — works less well if you want multi-model orchestration
The June 15, 2026 billing change moves Agent SDK / headless to separate API credits, which raises effective cost for heavy programmatic users

Effort to migrate: 5-10 days for typical teams. Significant workflow change.

Side-by-side

Dimension	Devin Local	Cursor	Claude Code
Surface	IDE (Devin Desktop)	IDE (Cursor)	Terminal (TUI)
Model integration	Devin family + ACP-compatible	Multi-model	Anthropic-first
Best feature	Lowest migration friction	Programmable SDK	Terminal-native
Pricing	Carries over from Windsurf	$20-40/mo+	Bundled with Claude Pro / Max
Autonomous agent	Devin Cloud (mature)	Limited (Bugbot, Composer)	Via background sub-agents
Migration effort from Cascade	1-3 days	3-7 days	5-10 days
Strategic risk	Cognition roadmap dependency	SpaceX acquisition pending	Anthropic single-vendor risk

How to choose

Stay on Devin Desktop with Devin Local if:

You’re happy with the current IDE and your team is productive
You use Devin Cloud for autonomous tasks
You want the lowest-effort migration

Switch to Cursor if:

You want to build custom tools that the agent can call
Your team has multi-model orchestration in production
You value the Cursor SDK’s programmability over Devin’s roadmap

Switch to Claude Code if:

You prefer terminal-first workflows
Your team is heavily Anthropic-aligned (or wants to be)
You run a lot of headless / CI / automated coding tasks

Migration timeline

For Devin Desktop users, here’s a realistic week-by-week plan:

This week (June 17-21):

Update Devin Desktop to the latest version
Switch one workflow to Devin Local in parallel with Cascade
Validate that rules and tool definitions carry over

Next week (June 22-28):

Migrate primary workflows to Devin Local
Re-tune any Cascade-specific prompts that behave differently
Roll back to Cascade only if you hit a blocker

Final week before EOL (June 29 - July 1):

All workflows on Devin Local
Disable Cascade fallback
Document any remaining quirks for your team

Don’t wait until June 30 to start — Cognition support will be saturated and any unexpected issues will queue.

What if you want to wait and switch later?

Reasonable. Migrate to Devin Local now (the safest bet given the deadline), then re-evaluate in Q4 2026 once:

Cursor’s SpaceX acquisition closes (Q3 2026 expected) and the post-deal direction is clear
Claude Fable 5 export-control situation resolves and Claude Code’s frontier-model story stabilizes
Cursor’s reportedly in-development frontier model ships (timing unclear)
Devin Local has 3-6 months of production track record

The migration cost from Devin Local to Cursor or Claude Code in Q4 will be roughly the same as today, so there’s no urgency to over-think the choice this month.

Bottom line

For most Windsurf/Devin Desktop users, the right move is Devin Local. The migration is the lowest friction, you keep what’s working, and you get a 30% efficiency improvement plus sub-agent support. Switch to Cursor or Claude Code only if you have a specific reason — not because the Cascade EOL is a forcing function.

The bigger picture: the AI coding IDE market is consolidating fast. Devin Desktop, Cursor, and Claude Code are the three real contenders for the next 12-18 months. Pick the one that fits your team’s workflow, plan for one more major re-evaluation in late 2026, and don’t over-invest in any single IDE today.

Cascade EOL July 1, 2026: Migrate to Devin Local, Cursor, or Claude Code?

TL;DR

What’s actually changing

The three migration targets

Devin Local (stay on Devin Desktop)

Cursor (switch IDEs)

Claude Code (switch to terminal-first)

Side-by-side

How to choose

Migration timeline

What if you want to wait and switch later?

Bottom line

Related on andrew.ooo