Background Agents

Claude Code

✓

Anthropic's terminal-based AI coding agent with Claude Opus 4.7, /ultrareview, Routines, /ultraplan, and 80.9% SWE-bench

Claude Code is Anthropic's terminal-first coding agent (GA since May 2025) that lives in your terminal and handles full development workflows. As of Week 16 (April 2026, v2.1.105–v2.1.113), Claude Opus 4.7 is the default model on Max and Team Premium with a new 'xhigh' effort level (recommended for most coding work) selectable via an interactive /effort slider. Recent shipments include /ultrareview for parallel multi-agent code review in the cloud (public research preview from Week 17, v2.1.119), Routines on Claude Code on the web for templated cloud agents triggered by schedule, GitHub events, or API, /usage breakdown showing what is driving your limits, native per-platform CLI binaries replacing bundled JavaScript, /ultraplan (Week 15) for cloud plan mode from the terminal, the Monitor tool for streaming background events, /autofix-pr to enable PR auto-fix from the CLI, /team-onboarding for replayable teammate ramp-up guides, and session recap (/recap) showing what happened in unfocused terminals. Custom themes can be built from /theme or shipped via plugins. Agent Teams enable multiple Claude Code instances to collaborate on complex tasks. Integrates with VS Code and JetBrains IDEs and supports GitHub Actions for automated code review, issue resolution, and PR creation.

• ●Terminal-based AI agent (GA since May 2025) for full development workflows
• ●Claude Opus 4.7 default on Max and Team Premium (Week 16, April 2026) with new 'xhigh' effort level and interactive /effort slider
• ●80.9% SWE-bench Verified: highest autonomous coding benchmark score

Cursor

✓

The AI-native code editor with $1B+ ARR, 25+ models, and background agents on dedicated VMs

Cursor is an AI-powered code editor built on a fork of VS Code by Anysphere, valued at $29.3B with over $1B in annual recurring revenue and 1M+ daily active users. It integrates AI directly into editing workflows with Supermaven-powered Tab autocomplete (sub-100ms latency), Composer for multi-file editing, and Agent mode for autonomous tasks. Cursor supports 25+ AI models including Claude Opus 4.5, Sonnet 4.6, GPT-5.2, and Gemini 2.5 Pro, and runs up to 20 concurrent background agents on dedicated AWS VMs for parallel development workflows.

• ●Supermaven-powered Tab autocomplete with sub-100ms latency and multi-line predictions
• ●Agent Mode (Cmd+L): conversational AI that reads, writes, and modifies files across your project
• ●Composer (Cmd+I): multi-file generation and editing from natural language

Google Antigravity

Agentic development platform powered by Gemini for autonomous coding

Google Antigravity is an agentic development platform powered by Gemini 3, designed for an agent-first era where AI agents handle complex development tasks. It offers an Editor View for hands-on coding and a Manager Surface for orchestrating multiple agents asynchronously. Agents can plan and execute tasks across editor, terminal, and browser, with an artifacts system for deliverables like task lists, screenshots, and browser recordings. Antigravity connects to Google Cloud via MCP for data sources like BigQuery, Spanner, Cloud SQL, and AlloyDB.

• ●Autonomous agents that plan and execute across editor, terminal, browser
• ●Editor View for hands-on coding
• ●Manager Surface for orchestrating multiple agents asynchronously

Gemini 3 Pro (with 3.1 Pro Preview & 3 Flash)

✓

Google's Gemini 3-series: 3.1 Pro Preview flagship, 3 Flash at ~78% SWE-bench, and 1M token context

The Gemini 3 series is Google's latest multimodal model lineup. Gemini 3 Flash is the default model across Google products, delivering ~78% SWE-bench Verified and 90.4% GPQA at just $0.50/$3 per MTok, making it the best price-performance coding model available. Gemini 3.1 Pro Preview (released Feb 19, 2026, at $2/$12 per MTok) is the newest flagship. Gemini 3 Pro achieves 76.2% SWE-bench but is discontinuing March 2026. All models support 1M token context windows, multimodal input, and tool-grounded workflows for coding and planning.

• ●Gemini 3 Flash: ~78% SWE-bench at $0.50/$3 per MTok (best price-performance)
• ●Gemini 3.1 Pro Preview: latest flagship released Feb 19, 2026
• ●Gemini 3 Pro: 76.2% SWE-bench (discontinuing March 2026)

OpenCode

Open-source AI coding agent for the terminal with multi-provider support

OpenCode is an open-source AI coding agent built for the terminal. It helps developers write code via a terminal interface, desktop app, or IDE extension. OpenCode supports many LLM providers including Claude, GPT, Gemini, and local models, with GitHub Copilot and ChatGPT Plus login options. It features LSP integration, multi-session agents, and a privacy-first approach that does not store code or context. The CLI supports run commands for single tasks, a headless API server, and MCP management.

• ●Terminal interface and desktop app (beta)
• ●opencode run for single-task execution
• ●opencode serve for headless API

Warp

✓

AI-native terminal with Oz cloud agent orchestration, Warp Drive, and Terminal-Bench #1 performance

Warp is an AI-native terminal reimagined as an Agentic Development Environment, built in Rust for speed and reliability. With 700K+ active developers and $73M+ in funding, Warp offers Agent Mode (Pair and Dispatch modes for autonomous task execution), Warp Drive for team knowledge sharing, and the Oz cloud agent orchestration platform launched February 10, 2026. Oz enables parallel cloud agents with cron scheduling, Slack/GitHub/Linear triggers, and multi-repo support. Warp holds the #1 position on Terminal-Bench at 52% and achieves 75.8% on SWE-bench when paired with GPT-5.

• ●Agent Mode: multi-step autonomous task execution in the terminal
• ●Pair Mode: AI assists alongside you as you work
• ●Dispatch Mode: AI works independently on complex tasks