SpaceX just bought the most popular AI code editor on earth for $60 billion. Windsurf doesn't exist anymore. And the majority of developers now use three or more AI coding tools at the same time, because no single one does everything well.
If you're scanning this for a ranked list, you'll get one. But the list looks different than it did eight weeks ago, and the reasons why are more useful than the rankings themselves.
Here's what actually shifted.
The big structural change: you're not picking a tool anymore, you're assembling a stack
A year ago the question was "Copilot or Cursor?" Today, 59% of developers use three or more AI coding tools in parallel. Cursor for editing. Claude Code for architecture and debugging. Copilot for completions inside VS Code. A terminal agent for CI. Maybe Devin running autonomously in the background on a separate task entirely.
This is not indecision. It's rational specialization. These tools got good enough at different things that using one for everything means leaving performance on the table.
The other number worth knowing: developer trust in AI code output dropped to 29%, down from 40% in 2024. More adoption, less trust. That tension shapes every tool decision right now.
The tools, ranked by what matters in July 2026
1. Claude Code
What it is: A terminal-native coding agent from Anthropic. No IDE required. You describe what you want, it reads your codebase, writes code, runs tests, iterates.
What changed: Claude Fable 5 launched June 9, and it is genuinely better at long, multi-step coding tasks than previous models. Anthropic also published a detailed breakdown of how "effort levels" work in Claude Code on July 7, clarifying that effort controls thoroughness (how many files the agent reads, how many tests it runs) separately from model capability. That distinction matters because it means you can run expensive models cheaply on simple tasks, or push cheaper models harder on complex ones.
Why it's here: Highest developer satisfaction of any AI coding tool. 46% of developers in an April 2026 survey called it "most loved," more than double Cursor's 19%. A 91% customer satisfaction score. Usage grew 6x in under a year. These are not marketing numbers from Anthropic. They're from independent surveys.
The catch: Claude-only models. No multi-model support. If you want GPT or Gemini, you need a different tool.
Price: Included with Claude Pro ($20/mo), Max ($100/mo), or Team/Enterprise plans. API access for custom integrations.
2. Cursor
What it is: A VS Code fork with AI built into every surface. Autocomplete, multi-file editing, an autonomous agent, background tasks.
What changed: The big news is the SpaceX acquisition. On June 16, SpaceX (which had merged with xAI earlier in 2026) announced a $60 billion all-stock deal to acquire Cursor's parent company Anysphere. The deal is expected to close in Q3. Before that, Cursor was in talks to raise another $2 billion at a $50 billion valuation. This is a company that hit $100 million in annual revenue in January 2025 and crossed $2 billion by February 2026. That growth rate has no precedent in enterprise software.
On the product side, the /multitask mode (launched April) lets you break a large task into parallel subtasks, each handled by a separate agent. It's the closest thing to "manage a team of AI coders" that exists in an IDE today.
Why it's here: Still the best daily-driver IDE for most developers. Multi-model support (Claude, GPT, Gemini). Mature background agents. The acquisition raises questions about long-term independence, but the product today is excellent.
The catch: The SpaceX/xAI acquisition creates real uncertainty. Will Cursor stay model-agnostic or get locked into Grok? Nobody knows yet. Also, Cursor loses money on individual developer accounts. The credit system can get expensive for heavy users.
Price: Free tier, Pro at $20/mo, Business at $40/user/mo.
3. Devin Desktop (formerly Windsurf)
What it is: An agent command center with a full VS Code-compatible IDE inside. The most autonomous option on this list.
What changed: Almost everything. Cognition (which acquired the original Windsurf/Codeium) shipped an OTA update on June 2 that renamed the entire product to Devin Desktop. The local agent, previously called Cascade, was rewritten from scratch in Rust as "Devin Local" (30% more token-efficient, supports sub-agents). Cascade hits end-of-life July 1. Devin Cloud, the fully autonomous agent that runs on its own cloud machine, is now available starting at the $20 Pro tier. It used to be enterprise-only.
Cognition also raised over $1 billion at a $26 billion valuation in late May. Clients include Citi, Mercedes-Benz, Goldman Sachs, Dell, and branches of the U.S. military.
The product reframe is explicit: Devin Desktop is an agent management hub that happens to contain an IDE, not an IDE that happens to call an agent. You open it and the default view is a Kanban board of your running agents, not an editor window.
Why it's here: If you want to hand off entire tasks to an AI and come back to finished PRs, this is the strongest option. The new "Spaces" feature lets agents share context across sessions. Support for the Agent Client Protocol (ACP) means you can run other agents (Codex, Claude Agent, OpenCode) inside the same environment.
The catch: Pro went from $15 to $20 in March. The agent-first paradigm is disorienting if you're used to typing code yourself. And the rapid rebranding (Codeium to Windsurf to Devin Desktop in 18 months) is a lot of name changes for one product.
Price: Free tier, Pro at $20/mo, Teams at $30-40/user/mo, Enterprise custom.
4. GitHub Copilot
What it is: Microsoft's AI coding assistant, built into VS Code and the broader GitHub ecosystem.
What changed: The June 1 billing overhaul is the big one. Copilot moved from "all you can use within a premium quota" to explicit usage-based pricing. Pro+ costs $39/month and gives you 3,900 AI credits (one credit = one cent). Overages are metered. This is a meaningful change for heavy users who were previously getting unlimited completions.
On the feature side: agentic browser tools are now enabled by default, letting Copilot agents navigate web pages and validate web apps directly inside VS Code. Parallel chat sessions. Per-session cost visibility. The "Autopilot" mode is getting better at running multi-step tasks without manual steering.
Why it's here: Distribution. 90% of Fortune 100 companies use it. 20 million total users. 4.7 million paid subscribers. If your company already pays for GitHub Enterprise, Copilot is probably included. That matters more than feature comparisons for a lot of teams.
The catch: Lowest developer satisfaction among the top four. Only 9% called it "most loved" in the same April survey where Claude Code hit 46%. The usage-based billing surprised people. And the product still feels like it's catching up to Cursor and Claude Code on agentic capabilities.
Price: Pro at $10/mo, Pro+ at $39/mo, Business at $19/user/mo, Enterprise at $39/user/mo.
5. OpenAI Codex (GPT-5.3-Codex)
What it is: OpenAI's dedicated coding model, available through ChatGPT, the API, and increasingly through third-party IDEs.
What changed: The February launch of GPT-5.3-Codex combined frontier coding with stronger reasoning in one model. 25% faster than its predecessor. State-of-the-art on the standard coding benchmarks. The interesting claim: it was the first Codex model used to debug its own training and deployment pipeline.
Why it's here: If you're already paying for ChatGPT and want a coding assistant without switching IDEs, Codex works. It handles long-running tasks with interactive steering. And because Cursor, Windsurf/Devin, and others all offer GPT models in their model picker, you might be using Codex without realizing it.
The catch: No standalone IDE. OpenAI has been slower than Anthropic and Cognition to ship agentic coding features. The best Codex experience is through a third-party editor, which says something.
Price: Included with ChatGPT Plus ($20/mo) and higher tiers. API pricing separate.
Also worth watching
Augment Code launched two products in quick succession. "Intent" (April) is a post-IDE workspace for coordinating multiple agents on real codebases, with isolated git worktrees per agent. "Cosmos" (GA in June) is an operating system for agentic software development across the full lifecycle: triage, spec, implementation, review, testing, deployment. Both are ambitious bets that the IDE itself is the wrong unit of abstraction.
Amazon Q Developer remains enterprise-focused with strong AWS integration but hasn't shipped a major feature update this summer. If you're deep in AWS, it's worth evaluating. Otherwise, the tools above are further ahead on agentic capabilities.
Sourcegraph Cody ($9/mo Pro) and Tabnine ($12/mo Pro) are solid if you need something lightweight or have specific compliance requirements, but neither is pushing the frontier right now.
What this means for how you work
If you're using one tool, add a second. The data says most developers already have. Claude Code for complex reasoning and architecture, your IDE's built-in agent (Cursor, Copilot, Devin Desktop) for daily editing. They complement each other because they're good at genuinely different things, not because the market is fragmented.
The Cursor acquisition deserves your attention even if you don't use Cursor. A million-plus developers are on Cursor Pro. If SpaceX/xAI locks it into Grok models or changes the pricing, those developers move. Where they land reshapes the market. Having some familiarity with Devin Desktop or Claude Code now is cheap insurance regardless.
And the trust question is worth sitting with. AI coding tools are writing more code than ever, and developers trust the output less than they did two years ago. That's not a paradox. The tools got powerful enough that the failure modes got more consequential. The ones that win the next phase will make verification easier, not just generation faster.