The devtools category overtakes mid-tier SaaS — Cursor $1.2B, Claude $2.5B, and the agent-IDE budget absorbs what was JetBrains plus CI plus Copilot
Cursor reached $1.2B ARR. Claude $2.5B annualized. The developer-tool category is now larger than the mid-tier SaaS category that dominated 2018-2024 analyst decks. The migration is visible in the financials of every meaningful vendor. The structural story is what happens to the SaaS revenue pool the migration just drained.
The category-overtake data
Industry analysis as of May 2026 places Cursor at $1.2B ARR, Claude at $2.5B annualized run rate, and Devin/Cognition at $400M+ on the autonomous-engineering tier. The developer-tool category is now larger than the mid-tier SaaS category that dominated 2018-2024 enterprise software analyst decks.
What budget did the migration drain
The agent-IDE category absorbed budget from several preceding categories:
- JetBrains / IDE licensing. Developers who previously paid $150-$300/year per developer for premium IDE licenses now pay $20-$200/month per developer for agent-IDE subscriptions. Each individual increase, multiplied across thousands of developers per enterprise.
- Continuous integration / build infrastructure. Some CI workloads (smoke tests, code quality checks) have shifted into agent-IDE pre-commit hooks. The reduction is partial but visible.
- GitHub Pro / advanced features. Copilot's incorporation into the agent-IDE category captured the budget that would otherwise have flowed into the GitHub Pro / Enterprise tier.
- Junior developer compensation, indirectly. Teams that would have hired additional junior developers absorb agent-IDE subscriptions instead. The economic effect is visible in agent-IDE ARR but obscured in IT budgets that account for it differently.
The two-vendor market structure
The category has settled into the two-vendor split we mapped earlier: per-token pricing (Cursor, Claude Code, OpenAI Codex) vs subscription-bundled quota (Windsurf+Devin at $200/month Max, GitHub Copilot). Both pricing models scale at hyperscaler-software rates.
The agent-IDE category isn't a niche of developer tools. It IS the developer tools category. The pre-2024 SaaS layer that served the same developer audience has been replaced.
The MCP-protocol substrate compounds the moat
MCP server registry continues to expand at 100-150 new servers per month, with over 800 production servers indexed. The protocol's surface coverage compounds for every agent-IDE vendor that consumes MCP — meaning the tool-integration layer is genuinely portable across the two-vendor split. Vendor lock-in lives at the model-routing and IDE-UX layers, not at tool integration.
The Google + Anthropic alignment
Google's Flash-tier consolidation across the Gemini app, Search, Spark, and Antigravity is the agent-tier-default strategy at one layer; Anthropic's Claude Code at a comparable price tier serves the same workload. The structural picture is that the agent-IDE category absorbs the budget while the model-tier inside the category increasingly consolidates onto Flash-equivalent pricing across both Google and Anthropic.
The forward read
Through 2026 H2, expect the category to consolidate further. The Cursor / Windsurf+Devin / Claude Code / Copilot quadrant captures the bulk of the budget; new entrants face genuinely hard go-to-market math against ARR scale and MCP-server ecosystem maturity. The Q4 2026 / Q1 2027 watch is whether enterprise IT budgets explicitly recognize agent-IDE as a separate line item, or whether the category continues to absorb mid-tier SaaS spend without an accounting reorg.
Blink Blog — best AI coding agents 2026 → · MightyBot — best AI coding agents 2026 → · Developers Digest — AI coding tools pricing 2026 →