agentlist

Shortlist

Which coding agent to shortlist

Claude Code and Cursor lead this shortlist. Claude Code is the commercial terminal agent, with a run-rate above $2.5 billion stated by Anthropic in February 2026. Cursor is the editor with agents built in, with a run-rate above $4 billion reported by Forbes in June 2026. Codex CLI is the main open-source alternative, with more than 5 million weekly users reported by OpenAI in June 2026.

Quick pick

  • Top pick for the terminal

    Claude Code

    Anthropic's command-line agent. It works inside your repo.

  • Top pick for an IDE

    Cursor

    Cursor 3’s Agents Window runs many agents at once, local or cloud.

  • Best open-source option

    OpenAI Codex

    OpenAI's open-source command-line agent, with a cloud version.

Ordered shortlist

  1. 01
    Claude Code

    Commercial

    Coding agents

    Anthropic's command-line coding agent.

    $2.5B+ annual run-rateVerifiedAs of Feb 12, 2026Anthropic

    Company-wide run-rate about $81 billionVerified As of Oct 31, 2026

    Evidence notes
    • Evidence level: Verified, primary.
    • Terminal agent. The run-rate is reported, not a fetched company filing.
    • Anthropic says it more than doubled since January 1, and enterprise is over half of it. Anthropic’s whole-company run-rate was $81 billion at the end of October (CNBC, 2026-11-10, citing its updated IPO filing).
    • Company-wide run-rate about $81 billion. From Anthropic’s amended IPO filing, up $10 billion from September. It covers all of Anthropic, not Claude Code alone, and the filing says the figure is not independently verified.
    • 155,099 GitHub stars. GitHub star count from the 2026-11-09 snapshot.
    • 155,099 GitHub stars. GitHub snapshot. As of 2026-11-09.
  2. 02
    Cursor

    Commercial

    Agentic development environments

    An editor with an agents-first window and a command line.

    $4B+ annual run-rateReportedAs of Jun 8, 2026Forbes

    About $60B implied equity valueReported As of Nov 9, 2026

    Evidence notes
    • Evidence level: Reported.
    • Editor with agents. The run-rate and the equity figure are separate.
    • From a person familiar with the matter, up from $3 billion in late April (Bloomberg) and $2 billion in February. Cursor told Forbes about 75% comes from businesses. Cursor itself confirmed above $1 billion in November 2025.
    • About $60 billion implied equity. TechCrunch reported talks at $50 billion on 2026-04-17. That is an earlier, separate figure.
  3. 03
    OpenAI Codex

    Open source

    Coding agents

    OpenAI's open-source command-line agent, with a cloud version.

    5M+ weekly usersVerifiedAs of Jun 2, 2026OpenAI

    40M weekly users, combined with ChatGPT WorkReported As of Oct 7, 2026

    Evidence notes
    • Evidence level: Verified, primary.
    • Open-source CLI and a cloud product. One user figure for both.
    • Codex alone, per OpenAI. About 20% are knowledge workers. OpenAI has not published a Codex-only count since the July merge with ChatGPT Work, and does not break out Codex revenue.
    • 40 million weekly users, Codex and ChatGPT Work combined. A combined figure for two products, not a Codex count.
    • 134,441 GitHub stars. GitHub star count from the 2026-11-09 snapshot.
    • 134,441 GitHub stars. GitHub snapshot. As of 2026-11-09.
  4. 04
    GitHub Copilot

    Commercial

    Agentic development environments

    GitHub's coding agent inside your IDE.

    50 million total usersReportedAs of July 2026AI Business

    Evidence notes
    • Evidence level: Reported.
    • Default IDE agent. The 50 million user figure is a third-party estimate.
    • Placed by judgment, above its survey rank. An ARR estimate on the source page was not used.
    • Survey notes also record 4.7 million paid subscribers as of January 2026. The stored source page is an ARR article. An ARR dollar estimate on that page was not used.
  5. 05
    Devin

    Commercial

    Coding agents

    A hosted agent that takes a task and works in the cloud.

    $48B post-money, almost $900M run-rateVerifiedAs of Nov 9, 2026Cognition blog

    Evidence notes
    • Evidence level: Verified, primary.
    • Hosted agent from Cognition. $48 billion post-money and a run-rate near $900 million, both on its Series E post.
    • The survey fetched this page. The post date was not copied onto the card. Named customers on the survey row: NVIDIA, GE Aerospace, Citi, Mercedes-Benz, and Modal. Talk of a $70–80 billion valuation was not used.
  6. 06
    Factory

    Commercial

    Coding agents

    Droid agents on desktop, terminal and cloud, with any model.

    $5B valuationReportedAs of Sep 15, 2026Reuters

    About $60 million annualized revenue, March 2026Estimated As of 2026-03

    Evidence notes
    • Evidence level: Reported.
    • Droid runs on desktop, terminal and cloud with any model. $5 billion valuation (Reuters, Sep 2026). Plans from $20 a month.
    • Reuters reported that Factory tripled its valuation.
    • About $60 million annualized revenue, March 2026. Sacra’s estimate, up from $5 million at the end of 2025.
  7. 07
    OpenCode

    Open source

    Coding agents

    An open-source command-line coding agent.

    225,683 GitHub starsGitHubAs of Nov 9, 2026GitHub repository

    Evidence notes
    • Evidence level: GitHub snapshot.
    • Open-source CLI. Stars are the scale signal.
    • GitHub star count from the 2026-11-09 snapshot.
  8. 08
    OpenHands

    Open source

    Coding agents

    Open-source control center for running several coding agents yourself.

    93,250 GitHub starsGitHubAs of Nov 11, 2026GitHub repository

    Evidence notes
    • Evidence level: GitHub snapshot.
    • Open-source control center that runs Claude Code, Codex and other agents on your own machine. 93,250 stars.
    • GitHub star count on Nov 11, 2026. Stars show interest, not users.
  9. 09
    Gemini CLI

    Open source

    Coding agents

    Google's open-source command-line coding agent.

    108,746 GitHub starsGitHubAs of Nov 9, 2026GitHub repository

    Evidence notes
    • Evidence level: GitHub snapshot.
    • Open-source CLI. Antigravity is the commercial successor surface.
    • GitHub star count from the 2026-11-09 snapshot.
  10. 10
    Cline

    Open source

    Coding agents

    An open-source coding agent that runs inside your IDE.

    71,467 GitHub starsGitHubAs of Nov 9, 2026GitHub repository

    About $5M ARREstimated As of August 2025

    Evidence notes
    • Evidence level: GitHub snapshot.
    • IDE extension. Stars lead. The ARR estimate stays secondary.
    • GitHub star count from the 2026-11-09 snapshot.
    • About $5 million ARR. Third-party estimate, low confidence.
  11. 11
    Aider

    Open source

    Coding agents

    An open-source command-line pair programmer.

    50,042 GitHub starsGitHubAs of Nov 9, 2026GitHub repository

    Evidence notes
    • Evidence level: GitHub snapshot.
    • Open-source CLI. Stars are the scale signal.
    • GitHub star count from the 2026-11-09 snapshot.
  12. 12
    Google Antigravity

    Commercial

    Agentic development environments

    Google’s agent-first app, free for individuals.

    2.4M+ weekly active usersReportedAs of Jul 23, 2026Kalkine

    Evidence notes
    • Evidence level: Reported.
    • Commercial IDE with no sourced dollar figure.
    • Placed by judgment. Listed because Google’s catalog treats it as the successor to Gemini Code Assist surfaces.
    • As reported by Kalkine; the page could not be opened, so the figure is carried as reported. At Google I/O in May 2026 Google said millions of developers use Antigravity.

Compare on

BrandTypeSurfaceLicenseHeadline figure
Claude CodeCommercialCLINot open source$2.5B+ annual run-rate · Verified
CursorCommercialIDENot open source$4B+ annual run-rate · Reported
OpenAI CodexOpen sourceCLIApache-2.05M+ weekly users · Verified
GitHub CopilotCommercialIDE, with CLI and cloud surfacesNot open source50 million total users · Reported
DevinCommercialCloud agent and desktop appNot open source$48B post-money, almost $900M run-rate · Verified
FactoryCommercialDesktop, CLI and cloudNot open source$5B valuation · Reported
OpenCodeOpen sourceCLIMIT225,683 GitHub stars · GitHub
OpenHandsOpen sourceSelf-hosted control center for agentsMIT93,250 GitHub stars · GitHub
Gemini CLIOpen sourceCLIApache-2.0108,746 GitHub stars · GitHub
ClineOpen sourceIDE extensionApache-2.071,467 GitHub stars · GitHub
AiderOpen sourceCLIApache-2.050,042 GitHub stars · GitHub
Google AntigravityCommercialIDENot open source2.4M+ weekly active users · Reported

Not recorded in the research: model lock-in, dollar prices, repo-size handling, and enterprise controls.

Cloud and background agents

These run the job off the laptop. Devin and Factory are ranked in the table above. OpenAI Codex cloud uses the same user figure as Codex CLI and is not counted again.

    Code review

    Code review is a neighbour, not a ranked list on this page. The survey does not support a scored code-review shortlist here, so no brands are ordered in this section.

    How we picked

    • Order follows the coding-agents survey, one product per row.
    • GitHub Copilot and Google Antigravity are judgment placements and are labelled. Copilot’s survey rank was lower. Antigravity had no sourced figure.
    • Star counts are the 2026-11-09 GitHub snapshot. The snapshot badge is not a revenue check.
    • Devin and Factory are hosted agents and OpenHands is an open-source control center. All three joined the table in November 2026.

    Questions and answers

    Which AI coding agent is best?
    Our judgmentClaude Code and Cursor lead this shortlist. Claude Code is the terminal agent, with a run-rate above $2.5 billion stated by Anthropic in February 2026. Cursor is the editor with agents built in, with a run-rate above $4 billion reported by Forbes in June 2026. This is an evidence order, not a measured quality rank.
    Which coding agents are free?
    OpenCode (MIT), OpenHands (MIT), Gemini CLI (Apache-2.0), Cline (Apache-2.0), Aider (Apache-2.0), and Codex CLI (Apache-2.0) are open source. That prices the software, not the model. Dollar prices for Claude Code, Cursor, and Copilot were not recorded here.
    How do coding agents work?
    The twelve above read a repository and edit code from a CLI, an IDE, or an IDE extension. Devin and Factory run the job off the laptop in the cloud. Interface labels come from the survey subtype.
    Where do coding agents fail?
    Our judgmentThis page has no measured failure study. The weak spot in the evidence is Claude Code’s run-rate, which is a third-party summary of a company statement. A tracker estimate of $15.1 billion was not used.
    How do Claude Code, Cursor, and Codex compare?
    Our judgmentClaude Code and Cursor are the commercial pair, one a CLI and one an IDE. Codex is the open-source CLI plus a cloud product. OpenAI reported more than 5 million weekly users in June 2026. Spec comparisons already exist for Claude Code vs Codex CLI and Claude Code vs Cursor.

    Runners-up

    • Open Interpreter, SWE-agent, Goose, Continue, Qwen Code

      Named in the survey. No figure is attached here.

    • Zed, Warp, Kiro, Replit Agent, Coder

      Named runners. No figure is attached here.

    • GitHub Copilot coding agent

      Same product as GitHub Copilot in the table above.

    Neighbours

    • Cloud agents

      Hosted and background agents are a section on this page until they have their own page.

    • Code review

      This page does not rank code-review tools. The survey does not support a separate scored shortlist here.

    • App builders

      Lovable and other app builders are a different job.

    • Customer support

      Resolving a company’s customers is a different job from writing code.

    Homepage shortlistAgentic development environmentsCustomer supportLegalApp buildersMCPInsurance