Zed is a high-performance code editor built entirely from scratch in Rust, not a fork or extension of VS Code or any existing editor. Released as open source in January 2024 and reaching 1.0 in April 2026, Zed raised a $32M Series B from Sequoia in August 2025, signaling serious commercial backing behind what began as a performance-first engineering project. With 85,000+ GitHub stars it has become one of the fastest-growing editor projects in recent memory.
Key capabilities
GPU-accelerated native rendering - Zed renders at 120fps using a custom GPU pipeline written in Rust, eliminating the Electron overhead that affects VS Code, Cursor, and most AI coding tools. The result is near-instant startup, fluid scrolling, and zero UI jank even on large files.
Parallel AI agent threads - Zed allows multiple agent threads to run concurrently on separate files or tasks, so a developer can have one agent refactoring a module while another writes tests for a different file. This is a structural advantage over editors that serialize agent work into a single context window.
Agent Client Protocol (ACP) - Zed implements ACP, an open protocol for coordinating multiple AI agents within a single workflow. This positions Zed as a participant in emerging multi-agent standards rather than a closed ecosystem.
Broad model support - The editor connects to Claude, GPT-4o, Gemini, Llama, and DeepSeek, giving teams flexibility to choose or rotate models without switching tools.
Autonomy level
Zed operates at autonomy level 3. Agents can read files, make multi-file edits, run terminal commands, and coordinate across threads, but the developer retains review control over proposed changes. Zed does not auto-commit or deploy autonomously; it surfaces agent output for human acceptance before changes land in the working tree.
Strengths
- Not Electron-based: native Rust with 120fps GPU rendering delivers the fastest editor experience in its class
- Parallel agent threads allow simultaneous AI work across multiple files or concerns
- Agent Client Protocol (ACP) support for multi-agent coordination
- Dual GPL-3.0 / Apache-2.0 license with full source on GitHub
- 85,000+ stars and active community development
- Broad model support including Claude, GPT-4o, Gemini, Llama, and DeepSeek
- $32M Series B funding provides a credible path to long-term maintenance
Limitations
- Agent capabilities are less mature than purpose-built agentic tools such as Cursor or Windsurf
- Primarily a code editor with AI features layered on, not an agent-first product designed around autonomous task completion
- Windows support arrived later than macOS and Linux and may lag in polish
- No built-in browser automation or MCP support at the level of tools like Kilo Code
- Smaller extension ecosystem than VS Code-based editors