Rank Claude Code, Codex CLI, Cursor and 24 more coding agents. Start from our editorial ranking or build your own, then export a 16:9 image or share a link.
Select an agent for its full review, or a tier for our reasons.
One line on what each agent does, who it suits and what to watch for. Open an agent for pricing, setup and alternatives.
The default pick for most developers: mature agent workflows on real repositories.
CLI agents · From $20
Anthropic's terminal-first agent reads, edits, tests and refactors real repositories, and the same agent runs in IDEs, GitHub Actions and the Agent SDK.
CLI agents · Free tier · From $8 · Open source
OpenAI's open-source terminal agent works inside your existing repositories and connects to the wider Codex ecosystem.
AI IDEs · Free tier · From $20
A VS Code-based AI editor that balances Tab completion, inline edits and agents, with cloud agents and a CLI around it.
Strong choices that beat S tier for some setups, such as open source, GitHub or Google.
CLI agents · Free tier available · Open source
MIT-licensed, model-flexible terminal agent that also runs in a desktop app, IDEs and GitHub workflows.
Editor extensions · Free tier · From $10
Autocomplete, chat, code review, CLI help and agent mode across most editors and GitHub itself.
Editor extensions · Free tier available · Open source
Open-source agent for the editor and terminal with Plan/Act modes and approval for each edit and command.
AI IDEs · Free tier available
Google's agent-first platform runs agents across an IDE, terminal, browser and CLI, and reports their work as reviewable artifacts.
Cloud agents · Free tier available
Cognition's autonomous engineer plans, codes, tests and opens pull requests from parallel cloud agents.
CLI agents · From $20
Terminal-first agent that routes work across frontier models and can keep threads running on remote Orbs.
Capable agents with clear trade-offs in price, maturity or scope.
CLI agents · Free tier available
Grok's terminal agent combines plan approval, parallel subagents, headless scripts and ACP integration, with support for Claude Code skills and plugins.
AI IDEs · Free tier · From $20
Agent-first editor with local and cloud agents, continuing as Devin Desktop.
AI IDEs · Free tier · From $20
AWS's agentic IDE turns prompts into specs, designs, tasks, code and pull requests.
AI IDEs · Free tier · From $20
TraeCode is an AI coding environment with a free plan, paired with the TraeWork task assistant.
Cloud agents · Free tier available
Google's asynchronous agent hands GitHub tasks to cloud workers and returns pull requests.
Cloud agents
GitHub's cloud agent takes an issue or prompt, works in GitHub Actions and opens a pull request for review.
Cloud agents · From $20
Droids take engineering tasks across the terminal, IDEs, Slack, Jira, Linear and CI/CD.
Editor extensions · Free tier available · Open source
Open-source VS Code agent with custom modes, rules and any model provider.
Editor extensions · Free tier available · Open source
Open-source agent for VS Code, JetBrains, the CLI and the cloud, with broad model choice.
Editor extensions · From $100
Built for large codebases, with a Context Engine, IDE and CLI agents, and Cosmos cloud agents.
CLI agents · Free tier available · Open source
Free, open-source pair programmer that edits files in your local Git repository with almost any model.
CLI agents · Free
Google's agent-first terminal runs several asynchronous agents and replaces Gemini CLI for individual users.
Worth it for a specific need, model or ecosystem.
CLI agents · Free tier available · Open source
A minimal, MIT-licensed terminal agent built to be customized with skills, extensions and scriptable modes.
AI IDEs · Free tier · From $10 · Open source
Fast, open-source native editor with agent workflows, BYOK and support for external agents.
CLI agents · From $19 · Open source
Moonshot AI's terminal agent built around Kimi's long-context coding models, with an MIT-licensed CLI.
CLI agents · Free tier available · Open source
Open-source terminal agent optimized for Qwen Coder models, with subagents, MCP and headless mode.
Cloud agents · Free tier available · Open source
Open-source, model-agnostic platform for autonomous agents, run locally, self-hosted or in OpenHands Cloud.
Being replaced; not recommended for new setups.
CLI agents · Free tier available · Open source
Google's open-source terminal agent, which individual users now replace with Antigravity CLI.
This is an editorial ranking, not a benchmark. We place each agent using the six criteria below, based on the capabilities, pricing and limits documented on its aiidelist.com page. We haven't run reproducible task tests, so there are no numeric scores.
Real-world task success
Fixing bugs, building features and getting tests to pass.
Reliability and safety
Staying in scope, asking before risky actions, and how much review the output needs.
Codebase understanding
Working across many files and large repositories.
Autonomy
Planning, tool use and running without supervision.
Pricing and value
Free tiers, quotas and what real tasks cost.
Developer experience
Setup, CLI or editor ergonomics, and integrations.
Agents and models are separate things: the same agent can work very differently with another model, so weigh each tier against the models and budget you have. Last reviewed . Spotted something out of date? Tell us.
In our editorial ranking, Claude Code, Codex CLI, and Cursor are S tier. The best choice still depends on the models you use, your budget, and whether you prefer working in a terminal, an IDE or with cloud agents.
No. It is editorial opinion based on each agent's documented capabilities, pricing and limits. Benchmarks such as SWE-bench Pro and Terminal-Bench test specific model and agent setups, and their results change with the version and configuration.
Choose Make your own above. Drag agents between tiers on a computer, or tap an agent and then a tier on a phone. Then export a 16:9 PNG or copy a link to share it. Your list is saved in this browser.
No. A shared link shows a visitor's own ranking. We don't review shared lists, and search engines don't index them.
