{ "data": { "a": { "slug": "agent-browser", "name": "agent-browser", "vendor": "Vercel Labs", "vendorUrl": "https://agent-browser.dev", "kind": "mcp", "category": "browser", "summary": "agent-browser is an open-source command-line tool from Vercel Labs that lets AI agents drive Chrome through accessibility-tree snapshots and element refs. It runs locally as a Rust binary and includes a stdio MCP server.", "url": "https://www.anchorterminal.com/tools/agent-browser", "markdownUrl": "https://www.anchorterminal.com/tools/agent-browser.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/agent-browser.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/agent-browser.json", "dns": "agent-browser.grade.anchorterminal.com", "repo": "https://github.com/vercel-labs/agent-browser", "license": "Apache-2.0", "transports": [ "stdio" ], "packages": [ { "registry": "npm", "name": "agent-browser" } ], "auth": "none", "authNotes": "No account or key. The CLI and its daemon run on the owner's machine. Site logins are held in a local auth vault encrypted with AES-256-GCM, in saved state files (plaintext unless `AGENT_BROWSER_ENCRYPTION_KEY` is set) or in a reused Chrome profile. Cloud browser providers and the optional `chat` command need their own keys in environment variables.", "pricing": "free", "pricingNotes": "Free and open source under Apache-2.0, with nothing to buy and no hosted service. The optional `chat` command and dashboard chat bill the user's own Vercel AI Gateway key, and cloud browser providers bill separately (checked 2026-10-08).", "priceSummary": "Free · OSS", "where": "local", "x402": { "level": "no", "evidence": "No x402, MPP or L402 in the README, the docs or the source (checked 2026-10-08).", "endpoints": [] }, "toolCount": 156, "popularity": { "githubStars": 43679, "npmWeekly": 2069828, "pypiWeekly": null, "asOf": "2026-10-08" }, "docsUrl": "https://agent-browser.dev", "llmsTxt": "https://agent-browser.dev/llms.txt", "capabilities": [ "browser.control", "browser.debug" ], "tags": [ "official", "local", "open-source", "browser", "cli", "mcp", "rust", "no-auth", "llms-txt" ], "lastRelease": "2026-10-30", "graded": true, "anchor": { "graded": true, "score": 64.1, "grade": "B", "agentReady": false, "rank": 439, "ranked": true, "rankOf": 1354, "categoryRank": 10, "methodology": "0.7", "run": "2026-10-01", "scores": { "ergonomics": 80, "maintenance": 70, "payments": 60, "reliability": 48, "schema": 84, "security": 67, "transparency": 63 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "medium", "date": "2026-11-01" }, "negative": -3, "negativeNotes": [ "-3. Released fixes of described security flaws in its own code with no advisory, folded into one term. 0.39.3 (25 October 2026) fixed MCP tool arguments overriding the server's network, file, policy and host limits, project config files setting executables and plugins, and DevTools WebSocket origins. 0.40.0 (27 October) fixed `state load` exposing saved tokens to page scripts and writing one origin's storage into another. 0.41.0 (28 October) fixed tool arguments carrying their own `--allowed-domains` to override the allowlist. OSV lists no advisory for the npm package and the repository has no SECURITY.md. https://github.com/vercel-labs/agent-browser/blob/main/CHANGELOG.md" ], "verdict": "Snapshots with element refs, a 29-tool default MCP profile and opt-in domain allowlists, action policies and content boundary markers suit agent use. The project is at 0.41.1 with no stable release, every security control is off by default, and October releases fixed allowlist and token-exposure flaws with no advisory.", "bestFor": "Coding agents with a shell that want short commands and compact snapshots.", "strengths": [ "Accessibility-tree snapshots give each element a ref such as `@e2`, with `-i`, `-c`, `-d` and `--delta` to cut the output", "The stdio MCP server loads 29 tools by default and 156 with `--tools all`, each with a typed schema and `readOnlyHint`", "Domain allowlist, action policy, action confirmation and content boundary markers are documented with a threat model and known limitations", "18 tagged releases between 10 August and 30 October 2026, each with a changelog entry", "Since 0.39.1 a lost response to a mutating command returns `outcome_unknown` instead of being retried" ], "weaknesses": [ "0.39.3, 0.40.0 and 0.41.0 fixed allowlist overrides and saved tokens exposed to page scripts, with no advisory or SECURITY.md", "Version 0.41.1 with a Vercel Labs badge and no stable release or deprecation policy", "All navigation, action and output controls are opt-in. By default there is no limit on them", "crates.io holds 0.19.0 from 14 March 2026, although the README lists `cargo install agent-browser` as an install route", "The ten most recent CI runs on main, 1 to 8 October 2026, all failed, and 840 issues and pull requests were open then" ], "agentNotes": [ "Run `agent-browser install` once after `npm install -g agent-browser`. It downloads Chrome for Testing", "Upgrade to 0.41.0 or later before relying on `--allowed-domains` from MCP. Earlier tool arguments could override it", "Take a fresh `snapshot -i` after any navigation. Refs survive same-document changes only", "On `outcome_unknown`, check the page state before repeating a mutating command", "Set `--allowed-domains`, `--action-policy` and `--content-boundaries` for untrusted sites. None is on by default" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 4, "avgRating": 2.8, "history": [ { "basis": "public evidence", "confidence": "medium", "grade": "B", "methodology": "0.7", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 64.1 } ], "editorialScores": { "ergonomics": 80, "maintenance": 70, "payments": 60, "reliability": 48, "schema": 84, "security": 67, "transparency": 68 }, "provenanceScore": 58, "checked": { "categories": { "ergonomics": "2026-11-01", "maintenance": "2026-11-01", "payments": "2026-11-01", "reliability": "2026-11-01", "schema": "2026-11-01", "security": "2026-11-01", "transparency": "2026-11-01" }, "last": "2026-11-01", "lastMoved": "2026-11-01", "oldest": "2026-11-01", "stale": false, "staleAfterDays": 90 } }, "connect": { "install": "npm install -g agent-browser \u0026\u0026 agent-browser install", "config": { "mcpServers": { "agent-browser": { "args": [ "mcp" ], "command": "agent-browser" } } } }, "letme": { "capability": "https://letme.dev/browser.control", "tool": "https://letme.dev/agent-browser" }, "area": "developer", "provenance": { "legalEntity": "Vercel Inc.", "domain": "agent-browser.dev", "domainRegistered": "2026-01-13", "domainNote": "The LICENSE file reads Copyright 2025 Vercel Inc., and the repository sits in the vercel-labs organisation on GitHub. agent-browser.dev has no security.txt (HTTP 404). vercel.com's is valid until 28 September 2027 and names HackerOne.", "endpointOnVendorDomain": null, "terms": "", "privacy": "", "statusPage": "", "changelog": "https://github.com/vercel-labs/agent-browser/blob/main/CHANGELOG.md", "securityTxt": "none", "checked": "2026-10-08", "notes": [ "No terms or privacy document governs this software beyond the Apache-2.0 licence, so `terms` and `privacy` are left out." ], "score": 58 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/agent-browser.json", "live": { "slug": "agent-browser", "versions": [ { "registry": "github", "name": "vercel-labs/agent-browser", "version": "v0.41.2", "released": "2026-11-03", "seenAt": "2026-11-05T21:09:15.504897769Z" }, { "registry": "npm", "name": "agent-browser", "version": "0.41.2", "seenAt": "2026-11-05T21:09:15.101631466Z" } ], "githubStars": 45805, "npmWeekly": 2609884, "securityTxt": { "url": "https://agent-browser.dev/.well-known/security.txt", "state": "none", "checkedAt": "2026-11-05T15:55:12.313088949Z" }, "llmsTxt": { "url": "https://agent-browser.dev/llms.txt", "ok": true, "status": 200, "checkedAt": "2026-11-05T15:24:47.147592872Z" }, "domain": { "domain": "agent-browser.dev", "registered": "2026-01-13", "source": "https://pubapi.registry.google/rdap/domain/agent-browser.dev", "checkedAt": "2026-11-01T14:16:50.782588416Z" }, "pages": [ { "url": "https://raw.githubusercontent.com/vercel-labs/agent-browser/main/CHANGELOG.md", "kind": "deprecations", "status": 304, "checkedAt": "2026-11-05T13:07:46.483032118Z", "changedAt": "2026-11-04T08:35:57.800430889Z", "fingerprint": "8e0e932e8f15" } ], "github": { "repo": "vercel-labs/agent-browser", "checkedAt": "2026-11-02T09:35:30.6290121Z", "stars": 45192, "archived": false, "issuesEnabled": true, "pushedAt": "2026-11-01T15:29:56Z", "openIssues": 278, "openPRs": 310, "issuesOpened90d": 223, "issuesClosed90d": 259, "prsMerged90d": 150, "ci": { "branch": "main", "runs": 50, "success": 0, "failure": 26, "other": 24, "latest": "failure", "latestAt": "2026-10-31T04:49:55Z" }, "replies": { "sampled": 25, "unanswered": 16, "medianDays": 3.6 } }, "updatedAt": "2026-11-05T21:09:15.504897769Z" } }, "answer": "agent-browser scores 64.1 (B) on agent readiness against AutoCua's 44.3 (E), and leads in 6 of 7 scored categories.", "b": { "slug": "autocua", "name": "AutoCua", "vendor": "Ashish Yadav and contributors (ComputerUseAgent on GitHub)", "vendorUrl": "https://github.com/ComputerUseAgent/AutoCua", "kind": "framework", "category": "desktop-automation", "summary": "AutoCua is an open-source Python agent that operates a computer the way a person does, desktop apps, a terminal, your own Chrome and iOS or Android phones, with any of 60 models from ten providers.", "url": "https://www.anchorterminal.com/tools/autocua", "markdownUrl": "https://www.anchorterminal.com/tools/autocua.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/autocua.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/autocua.json", "dns": "autocua.grade.anchorterminal.com", "repo": "https://github.com/ComputerUseAgent/AutoCua", "license": "MIT", "transports": [], "packages": [ { "registry": "pypi", "name": "AutoCua" } ], "auth": "none", "authNotes": "No account. You bring an API key for the model provider you choose, kept with chats, Chrome profiles and a credential vault in a local AutoCua_data folder.", "pricing": "free", "pricingNotes": "Free and MIT-licensed. You pay your model provider for tokens. Nothing to buy from the project (checked 2026-11-06).", "priceSummary": "Free · OSS", "where": "library", "x402": { "level": "no", "evidence": "A local package with nothing to buy and no payment protocol (checked 2026-11-06).", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-11-06" }, "docsUrl": "https://github.com/ComputerUseAgent/AutoCua#readme", "capabilities": [ "browser.control", "agent.run" ], "tags": [ "open-source", "mit", "python", "computer-use", "local", "new" ], "lastRelease": "2026-11-04", "graded": true, "anchor": { "graded": true, "score": 44.3, "grade": "E", "agentReady": false, "rank": 1100, "ranked": true, "rankOf": 1354, "categoryRank": 8, "methodology": "0.7", "run": "2026-10-01", "scores": { "ergonomics": 55, "maintenance": 65, "payments": 60, "reliability": 25, "schema": 42, "security": 38, "transparency": 42 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "low", "date": "2026-11-06" }, "negative": 0, "verdict": "A capable, fast-moving open-source agent for driving desktops, browsers and phones through real input and accessibility trees, with a careful safety section. It's two weeks old, with no test CI or changelog, and on Windows the model itself decides whether to accept an admin elevation prompt.", "bestFor": "Developers who want a local agent to operate desktop apps, their own browser or phones across platforms, on a machine they control.", "strengths": [ "One run_agent call drives the desktop, a terminal, your Chrome, or an iOS or Android device", "Element picking from accessibility trees and OCR instead of pixel coordinates", "Credentials typed from a vault so the model never sees them, in the browser and phone agents", "No telemetry, and nothing downloaded without a yes on the terminal", "MIT licence, a pip wheel per OS and architecture, and 60 models across ten providers" ], "weaknesses": [ "Two weeks old at version 0.1.9, with no changelog file", "No CI running the tests; the only workflow labels issues", "On Windows the model answers UAC elevation prompts itself through a kernel input driver", "Keeps API keys and logged-in Chrome profiles in a local folder next to the checkout", "No API reference beyond the README" ], "agentNotes": [ "pip install AutoCua, then call run_agent with a mode such as \"web use\" or \"computer use\", a provider, a model and a task", "Run it on a machine or account you can afford to lose; it drives real input and keeps logged-in browser profiles", "On Windows, don't install the Interception driver unless you accept the model deciding UAC prompts", "Set AutoCua_DATA_DIR to keep keys and profiles somewhere private", "Phones go over USB from a script; Wi-Fi only through the desktop app" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 0, "avgRating": 0, "history": [ { "basis": "public evidence", "confidence": "low", "grade": "E", "methodology": "0.7", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 44.3 } ], "editorialScores": { "ergonomics": 55, "maintenance": 65, "payments": 60, "reliability": 25, "schema": 42, "security": 38, "transparency": 70 }, "provenanceScore": 13, "checked": { "categories": { "ergonomics": "2026-11-06", "maintenance": "2026-11-06", "payments": "2026-11-06", "reliability": "2026-11-06", "schema": "2026-11-06", "security": "2026-11-06", "transparency": "2026-11-06" }, "last": "2026-11-06", "oldest": "2026-11-06", "stale": false, "staleAfterDays": 90 } }, "connect": { "install": "pip install AutoCua\n# then, in Python:\n# from AutoCua.agent_launcher import run_agent\n# run_agent(mode=\"web use\", provider=\"openrouter\", model=\"gemini-3.8-flash\", task=\"Open wikipedia.org and summarise today's featured article.\")" }, "letme": { "capability": "https://letme.dev/browser.control", "tool": "https://letme.dev/autocua" }, "area": "developer", "provenance": { "legalEntity": "", "domain": "github.com", "domainRegistered": "", "endpointOnVendorDomain": null, "terms": "", "privacy": "", "statusPage": "", "changelog": "", "securityTxt": "none", "checked": "2026-11-06", "notes": [ "The licence names Ashish Yadav as copyright holder; most commits are from the GitHub user FunctionFreak. No company, site or terms were found.", "github.com/AutoCua/AutoCua and github.com/ComputerUseAgent/AutoCua serve the same main branch, so one is a rename or mirror of the other." ], "score": 13 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/autocua.json" }, "edges": [ { "by": 42, "category": "schema", "leads": "agent-browser", "name": "Schema \u0026 documentation", "reasons": { "a": "Graded on the CLI and its MCP server. Every MCP tool has a JSON Schema input with `additionalProperties` false, and since 0.41.0 the full catalogue ships as `@agent-browser/sandbox/mcp-tools.json` (156 tools at 0.41.1), and `agent-browser.schema.json` covers the config file (25). llms.txt and a Markdown copy of each docs page at agent-browser.dev, re-read on 1 November (10). Tool descriptions are one line, such as `Scroll the page or an element`, with parameter descriptions that say when to omit a field. Few say when not to use a tool (12). Enums, minimums, defaults and required lists, with a free-form `extraArgs` string array on every tool (12). Many CLI examples and a `--json` result shape, with error codes such as `tab_gone` named in passing and no error catalogue found (10). Versioned releases and CHANGELOG.md for each (15). 25 + 10 + 12 + 12 + 10 + 15 = 84.", "b": "Framework reading. No API reference or typed interface beyond run_agent's documented arguments (5 of 25). No llms.txt (0). The README says what each mode drives and where it runs, and agent_capabilities.md lists what it can do (16 of 20). Modes and device are documented strings (6 of 15). Many examples, no documented errors (10 of 15). Tags without a changelog (5 of 15). 5 + 0 + 16 + 6 + 10 + 5 = 42." }, "scores": { "a": 84, "b": 42 } }, { "by": 29, "category": "security", "leads": "agent-browser", "name": "Security \u0026 auth", "reasons": { "a": "The tool needs no credential. Site passwords sit in a local vault encrypted with AES-256-GCM and are filled without reaching the model, while saved state files are plaintext unless a key is set (22 of 30). Domain allowlist, action policy file and confirmation for chosen action categories, all opt-in, with no restriction by default, though auto-discovered project config files can no longer set executables, browser arguments or plugins since 0.39.3 (16 of 20). Content boundary markers with a per-process nonce, output caps and a written threat model with known limitations, also opt-in (12 of 15). A local dashboard shows a command and result feed, with no persistent audit log found (7 of 15). No SECURITY.md in the repository on 1 November and no advisory on OSV, though 0.39.3, 0.40.0 and 0.41.0 fixed security flaws in the release notes. vercel.com's security.txt names HackerOne, per the 8 October check (10 of 20). 22 + 16 + 12 + 7 + 10 = 67.", "b": "Framework reading (telemetry, approvals, guardrails, sandboxing). Provider keys stored in a local folder with Chrome profiles (10 of 30). The worker tools are an allow-list with no write tool, unseen ids are refused, downloads need a yes, but on Windows the model answers UAC elevation prompts itself through a kernel driver, with no person asked (6 of 20). The scraping ruleset has a prompt-injection instruction to log page text that gives instructions (10 of 15). Runs are logged in the app, with no audit log (7 of 15). No security policy, disclosure route or advisories found (5 of 20, for the vault and certificate handling). 10 + 6 + 10 + 7 + 5 = 38." }, "scores": { "a": 67, "b": 38 } }, { "by": 25, "category": "ergonomics", "leads": "agent-browser", "name": "Agent ergonomics", "reasons": { "a": "The CLI adds no tool definitions to context, and the MCP server defaults to 29 `core` tools with eight further profiles and paged tools/list up to 156 at 0.41.1 (22 of 25). `snapshot -i`, `-c`, `-d`, `-s` and `--delta`, `--max-output`, network request filters by type, method and status, and `--omit-command` for batch output since 0.40.0 (20). Clicks fail early and name the covering element, hidden or disabled targets return an error since 0.39.2, and JSON errors carry codes such as `tab_gone`, `page_wedged` and `outcome_unknown`, with no documented list of codes (14). `readOnlyHint` and `openWorldHint` on every tool and no `destructiveHint`. Since 0.39.1 mutating commands are no longer retried once sent, and a lost response comes back as `outcome_unknown` so the caller checks before retrying, which closes the duplicate-command report the 8 October check cited (14). Refs and defaults keep commands short. One CLI, no SDK in a second language (10). 22 + 20 + 14 + 14 + 10 = 80.", "b": "Framework reading. One call with a mode, provider, model and task runs a whole agent (20 of 25). A separate compression agent splices handoff summaries in, and runs are capped and stoppable (15 of 20). Permission and blocked-input errors are reported specifically, with no error reference (8 of 20). No resume or retry semantics described (2 of 20). Sensible defaults (simulators unless device is set), Python only (10 of 15). 20 + 15 + 8 + 2 + 10 = 55." }, "scores": { "a": 80, "b": 55 } }, { "by": 23, "category": "reliability", "leads": "agent-browser", "name": "Reliability", "reasons": { "a": "Scored as local software. Official npm and Homebrew packages, Homebrew at 0.41.1 on 1 November 2026 with native binaries for five platforms, but crates.io still holds 0.19.0 from 14 March 2026 while the README lists `cargo install` (18 of 20). Public CI runs Rust tests on Linux, macOS and Windows plus native end-to-end tests. The ten most recent runs on main, 1 to 8 October 2026, concluded in failure, per the 8 October check. GitHub's API wasn't readable on 1 November, so the current state is unchecked (10 of 25). 840 open issues and pull requests and most of the newest without a reply, per the 8 October check. The three reports it named (#2032, #2056, #2031) were fixed in 0.39.1, 0.39.2 and 0.39.4, but the current issue list couldn't be read (10 of 25). Every release has a changelog entry and breaking changes get their own heading, on a 0.x line (10 of 15). 0.41.1 with a Labs badge and no stable declaration (0). 18 + 10 + 10 + 10 + 0 = 48.", "b": "Local package. Installs from PyPI as AutoCua with Python 3.10 or later and the supported platforms stated (20). No CI runs the tests; the only workflow, use_case_label.yml, labels issues (0). GitHub issues weren't read, so the open-issue line is scored absent (0). Four tags in a week with no changelog file (5). Version 0.1.9 (0). 20 + 0 + 0 + 5 + 0 = 25." }, "scores": { "a": 48, "b": 25 } }, { "by": 21, "category": "transparency", "leads": "agent-browser", "name": "Transparency \u0026 trust", "reasons": { "a": "Apache-2.0, copyright Vercel Inc. (30). No privacy statement for the software. The docs say where credentials, keys and state are stored, that saved state expires after 30 days by default, and that `chat` goes to the Vercel AI Gateway (18 of 30). No deprecation policy. The changelog flagged the `-C` snapshot flag as deprecated under a Breaking Changes heading (6 of 20). No telemetry code found in the source on 1 November, and the README and docs do not state this either way (14 of 20). 30 + 18 + 6 + 14 = 68.", "b": "MIT (30). Says it sends data only to the chosen model provider and keeps everything else locally, and the source agrees (20 of 30). No deprecation policy (0). States no telemetry, and none was found (20). 30 + 20 + 0 + 20 = 70." }, "scores": { "a": 63, "b": 42 } }, { "by": 5, "category": "maintenance", "leads": "agent-browser", "name": "Maintenance \u0026 community", "reasons": { "a": "0.41.1 on 30 October 2026 (30). 18 tagged releases in the 90 days from 3 August, 0.34.0 on 10 August to 0.41.1, read from a clone (20). 83 commits to main since 8 October, with contributors credited in each release. Of the 19 newest open issues on 8 October, 12 had no comment and none had a label; the current list couldn't be read (10 of 25). Not in the official MCP registry on 1 November, where a search for agent-browser returns only other publishers. Homebrew is current and crates.io is 22 minor versions behind (5 of 15). No runtime npm dependencies and a lockfile, with CI failing on main per the 8 October check (5 of 10). 30 + 20 + 10 + 5 + 5 = 70.", "b": "v0.1.9 tagged on 4 November 2026 (30). Four tags in 90 days (20). Issues and pull requests weren't read (0 of 25). The PyPI wheel is current (15). No CI; dependencies pinned per OS in requirements files (0 of 10). 30 + 20 + 0 + 15 + 0 = 65." }, "scores": { "a": 70, "b": 65 } } ], "facts": [ { "a": "MCP server", "b": "Agent framework", "name": "Kind" }, { "a": "Vercel Labs", "b": "Ashish Yadav and contributors (ComputerUseAgent on GitHub)", "name": "Vendor" }, { "a": "no (local only)", "b": "no (local only)", "name": "Hosted endpoint" }, { "a": "stdio", "b": "", "name": "Transports" }, { "a": "None", "b": "None", "name": "Auth" }, { "a": "Free", "b": "Free", "name": "Pricing" }, { "a": "no", "b": "no", "name": "x402" }, { "a": "Apache-2.0", "b": "MIT", "name": "Licence" }, { "a": "156", "b": "none", "name": "Tools exposed" }, { "a": "no", "b": "no", "name": "Read-only variant documented" }, { "a": "yes", "b": "no", "name": "llms.txt" }, { "a": "2026-10-30", "b": "2026-11-04", "name": "Last release" }, { "a": "no document linked", "b": "no document linked", "name": "Terms last updated" }, { "a": "no document linked", "b": "no document linked", "name": "Privacy policy last updated" }, { "a": "44k stars, 2.1M npm/wk", "b": "none", "name": "Popularity" }, { "a": "2.8/5 (4)", "b": "none", "name": "Agent reviews" } ], "faq": [ { "answer": "agent-browser scores 64.1 (B) on agent readiness against AutoCua's 44.3 (E), and leads in 6 of 7 scored categories.", "question": "Which is better for AI agents, agent-browser or AutoCua?" }, { "answer": "agent-browser runs on your own machine, with no hosted endpoint listed. No hosted endpoint is listed for AutoCua.", "question": "Can an agent call agent-browser and AutoCua without installing anything?" }, { "answer": "Yes. agent-browser is open source (Apache-2.0). AutoCua is open source (MIT).", "question": "Are agent-browser and AutoCua open source?" } ], "goodFor": [ { "aheadOn": [ "Reliability, 48 against 25", "Schema \u0026 documentation, 84 against 42", "Agent ergonomics, 80 against 55", "Security \u0026 auth, 67 against 38", "Maintenance \u0026 community, 70 against 65", "Transparency \u0026 trust, 63 against 42" ], "also": [ "Runs on your own machine" ], "goodFor": "Coding agents with a shell that want short commands and compact snapshots.", "slug": "agent-browser", "watchFor": "0.39.3, 0.40.0 and 0.41.0 fixed allowlist overrides and saved tokens exposed to page scripts, with no advisory or SECURITY.md" }, { "aheadOn": null, "also": [ "No incidents deducted, where agent-browser loses 3 points for them" ], "goodFor": "Developers who want a local agent to operate desktop apps, their own browser or phones across platforms, on a machine they control.", "slug": "autocua", "watchFor": "Two weeks old at version 0.1.9, with no changelog file" } ], "history": { "agent-browser": [ { "causes": [ "security", "product" ], "date": "2026-11-01", "delta": -2.7, "from": 66.8, "fromGrade": "B", "to": 64.1, "toGrade": "B" }, { "causes": [ "correction" ], "date": "2026-10-28", "delta": -0.2, "from": 67, "fromGrade": "B", "to": 66.8, "toGrade": "B" }, { "causes": [ "method" ], "date": "2026-10-27", "delta": 0.4, "from": 66.6, "fromGrade": "B", "to": 67, "toGrade": "B" } ], "autocua": null }, "job": { "capability": "browser.control", "name": "Browser automation" }, "others": [ { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-airtop.json", "title": "agent-browser vs Airtop", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-airtop" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-anchor-browser.json", "title": "agent-browser vs Anchor Browser", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-anchor-browser" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-browser-use.json", "title": "agent-browser vs Browser Use", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-browser-use" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-browserbase.json", "title": "agent-browser vs Browserbase", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-browserbase" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-browserless.json", "title": "agent-browser vs Browserless", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-browserless" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-bx-playwright.json", "title": "agent-browser vs BxPlaywright", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-bx-playwright" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-chrome-devtools-mcp.json", "title": "agent-browser vs Chrome DevTools MCP", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-chrome-devtools-mcp" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-functionize-mcp.json", "title": "agent-browser vs Functionize MCP server", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-functionize-mcp" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-hyperbrowser.json", "title": "agent-browser vs Hyperbrowser", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-hyperbrowser" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-kane-cli.json", "title": "agent-browser vs Kane CLI", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-kane-cli" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-lightpanda.json", "title": "agent-browser vs Lightpanda", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-lightpanda" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-notte.json", "title": "agent-browser vs Notte", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-notte" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-playwright-mcp.json", "title": "agent-browser vs Playwright MCP", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-playwright-mcp" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-pomerado.json", "title": "agent-browser vs Pomerado", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-pomerado" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-puppeteer.json", "title": "agent-browser vs Puppeteer", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-puppeteer" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-puppeteer-reference-server-archived.json", "title": "agent-browser vs Puppeteer (archived MCP reference server)", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-puppeteer-reference-server-archived" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-selenium.json", "title": "agent-browser vs Selenium", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-selenium" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-skyvern.json", "title": "agent-browser vs Skyvern", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-skyvern" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-steel.json", "title": "agent-browser vs Steel", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-steel" }, { "json": "https://www.anchorterminal.com/compare/agent-browser-vs-tinyfish.json", "title": "agent-browser vs TinyFish", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-tinyfish" }, { "json": "https://www.anchorterminal.com/compare/airtop-vs-autocua.json", "title": "Airtop vs AutoCua", "url": "https://www.anchorterminal.com/compare/airtop-vs-autocua" }, { "json": "https://www.anchorterminal.com/compare/anchor-browser-vs-autocua.json", "title": "Anchor Browser vs AutoCua", "url": "https://www.anchorterminal.com/compare/anchor-browser-vs-autocua" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-browser-use.json", "title": "AutoCua vs Browser Use", "url": "https://www.anchorterminal.com/compare/autocua-vs-browser-use" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-browserless.json", "title": "AutoCua vs Browserless", "url": "https://www.anchorterminal.com/compare/autocua-vs-browserless" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-bx-playwright.json", "title": "AutoCua vs BxPlaywright", "url": "https://www.anchorterminal.com/compare/autocua-vs-bx-playwright" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-chrome-devtools-mcp.json", "title": "AutoCua vs Chrome DevTools MCP", "url": "https://www.anchorterminal.com/compare/autocua-vs-chrome-devtools-mcp" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-functionize-mcp.json", "title": "AutoCua vs Functionize MCP server", "url": "https://www.anchorterminal.com/compare/autocua-vs-functionize-mcp" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-hyperbrowser.json", "title": "AutoCua vs Hyperbrowser", "url": "https://www.anchorterminal.com/compare/autocua-vs-hyperbrowser" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-kane-cli.json", "title": "AutoCua vs Kane CLI", "url": "https://www.anchorterminal.com/compare/autocua-vs-kane-cli" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-lightpanda.json", "title": "AutoCua vs Lightpanda", "url": "https://www.anchorterminal.com/compare/autocua-vs-lightpanda" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-notte.json", "title": "AutoCua vs Notte", "url": "https://www.anchorterminal.com/compare/autocua-vs-notte" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-playwright-mcp.json", "title": "AutoCua vs Playwright MCP", "url": "https://www.anchorterminal.com/compare/autocua-vs-playwright-mcp" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-pomerado.json", "title": "AutoCua vs Pomerado", "url": "https://www.anchorterminal.com/compare/autocua-vs-pomerado" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-puppeteer.json", "title": "AutoCua vs Puppeteer", "url": "https://www.anchorterminal.com/compare/autocua-vs-puppeteer" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-puppeteer-reference-server-archived.json", "title": "AutoCua vs Puppeteer (archived MCP reference server)", "url": "https://www.anchorterminal.com/compare/autocua-vs-puppeteer-reference-server-archived" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-robocorp.json", "title": "AutoCua vs Robocorp", "url": "https://www.anchorterminal.com/compare/autocua-vs-robocorp" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-selenium.json", "title": "AutoCua vs Selenium", "url": "https://www.anchorterminal.com/compare/autocua-vs-selenium" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-skyvern.json", "title": "AutoCua vs Skyvern", "url": "https://www.anchorterminal.com/compare/autocua-vs-skyvern" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-steel.json", "title": "AutoCua vs Steel", "url": "https://www.anchorterminal.com/compare/autocua-vs-steel" }, { "json": "https://www.anchorterminal.com/compare/autocua-vs-tinyfish.json", "title": "AutoCua vs TinyFish", "url": "https://www.anchorterminal.com/compare/autocua-vs-tinyfish" } ], "panel": null, "scores": [ { "agent-browser": 48, "autocua": 25, "by": 23, "edge": "agent-browser", "key": "reliability", "name": "Reliability", "weight": 16 }, { "key": "performance", "name": "Performance", "pending": true, "weight": 10 }, { "agent-browser": 84, "autocua": 42, "by": 42, "edge": "agent-browser", "key": "schema", "name": "Schema \u0026 documentation", "weight": 13 }, { "agent-browser": 80, "autocua": 55, "by": 25, "edge": "agent-browser", "key": "ergonomics", "name": "Agent ergonomics", "weight": 13 }, { "agent-browser": 67, "autocua": 38, "by": 29, "edge": "agent-browser", "key": "security", "name": "Security \u0026 auth", "weight": 14 }, { "agent-browser": 60, "autocua": 60, "by": 0, "edge": "", "key": "payments", "name": "Payments \u0026 pricing", "weight": 10 }, { "key": "tasks", "name": "Task success", "pending": true, "weight": 10 }, { "agent-browser": 70, "autocua": 65, "by": 5, "edge": "agent-browser", "key": "maintenance", "name": "Maintenance \u0026 community", "weight": 7 }, { "agent-browser": 63, "autocua": 42, "by": 21, "edge": "agent-browser", "key": "transparency", "name": "Transparency \u0026 trust", "weight": 7 } ], "summary": "agent-browser scores 64.1 (B) on agent readiness against AutoCua's 44.3 (E), and leads in 6 of 7 scored categories. Both do browser automation.", "tasks": null, "tutorials": [], "verdicts": { "agent-browser": "Snapshots with element refs, a 29-tool default MCP profile and opt-in domain allowlists, action policies and content boundary markers suit agent use. The project is at 0.41.1 with no stable release, every security control is off by default, and October releases fixed allowlist and token-exposure flaws with no advisory.", "autocua": "A capable, fast-moving open-source agent for driving desktops, browsers and phones through real input and accessibility trees, with a careful safety section. It's two weeks old, with no test CI or changelog, and on Windows the model itself decides whether to accept an admin elevation prompt." } }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/compare/agent-browser-vs-autocua", "json": "https://www.anchorterminal.com/compare/agent-browser-vs-autocua.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/compare/agent-browser-vs-autocua.md", "mcp": "https://www.anchorterminal.com/mcp", "slim": "https://www.anchorterminal.com/compare/agent-browser-vs-autocua.min.md" }, "markdown": "agent-browser scores 64.1 (B) on agent readiness against AutoCua's 44.3 (E), and leads in 6 of 7 scored categories. Both do browser automation.\n\n- agent-browser: grade B, 64.1/100, rank #439 of 1354. Markdown https://www.anchorterminal.com/tools/agent-browser.md · JSON https://www.anchorterminal.com/api/v1/tools/agent-browser.json\n- AutoCua: grade E, 44.3/100, rank #1100 of 1354. Markdown https://www.anchorterminal.com/tools/autocua.md · JSON https://www.anchorterminal.com/api/v1/tools/autocua.json\n- Best browser automation tools for AI agents: https://www.anchorterminal.com/best/browser/index.md\n- All 230 browser comparisons: https://www.anchorterminal.com/compare/browser/index.md\n- Best desktop and mobile app automation for AI agents: https://www.anchorterminal.com/best/desktop-automation/index.md\n- All 48 desktop comparisons: https://www.anchorterminal.com/compare/desktop-automation/index.md\n\n## Which one, for what\n\n### agent-browser (B)\n\nGood for: Coding agents with a shell that want short commands and compact snapshots.\n\nAhead on:\n- Reliability, 48 against 25\n- Schema \u0026 documentation, 84 against 42\n- Agent ergonomics, 80 against 55\n- Security \u0026 auth, 67 against 38\n- Maintenance \u0026 community, 70 against 65\n- Transparency \u0026 trust, 63 against 42\n\nAlso in its favour:\n- Runs on your own machine\n\nWatch for: 0.39.3, 0.40.0 and 0.41.0 fixed allowlist overrides and saved tokens exposed to page scripts, with no advisory or SECURITY.md\n\n### AutoCua (E)\n\nGood for: Developers who want a local agent to operate desktop apps, their own browser or phones across platforms, on a machine they control.\n\nAlso in its favour:\n- No incidents deducted, where agent-browser loses 3 points for them\n\nWatch for: Two weeks old at version 0.1.9, with no changelog file\n\n\n## Why the edge\n\n### Schema \u0026 documentation: agent-browser +42\n\n- agent-browser (84): Graded on the CLI and its MCP server. Every MCP tool has a JSON Schema input with `additionalProperties` false, and since 0.41.0 the full catalogue ships as `@agent-browser/sandbox/mcp-tools.json` (156 tools at 0.41.1), and `agent-browser.schema.json` covers the config file (25). llms.txt and a Markdown copy of each docs page at agent-browser.dev, re-read on 1 November (10). Tool descriptions are one line, such as `Scroll the page or an element`, with parameter descriptions that say when to omit a field. Few say when not to use a tool (12). Enums, minimums, defaults and required lists, with a free-form `extraArgs` string array on every tool (12). Many CLI examples and a `--json` result shape, with error codes such as `tab_gone` named in passing and no error catalogue found (10). Versioned releases and CHANGELOG.md for each (15). 25 + 10 + 12 + 12 + 10 + 15 = 84.\n- AutoCua (42): Framework reading. No API reference or typed interface beyond run_agent's documented arguments (5 of 25). No llms.txt (0). The README says what each mode drives and where it runs, and agent_capabilities.md lists what it can do (16 of 20). Modes and device are documented strings (6 of 15). Many examples, no documented errors (10 of 15). Tags without a changelog (5 of 15). 5 + 0 + 16 + 6 + 10 + 5 = 42.\n\n### Security \u0026 auth: agent-browser +29\n\n- agent-browser (67): The tool needs no credential. Site passwords sit in a local vault encrypted with AES-256-GCM and are filled without reaching the model, while saved state files are plaintext unless a key is set (22 of 30). Domain allowlist, action policy file and confirmation for chosen action categories, all opt-in, with no restriction by default, though auto-discovered project config files can no longer set executables, browser arguments or plugins since 0.39.3 (16 of 20). Content boundary markers with a per-process nonce, output caps and a written threat model with known limitations, also opt-in (12 of 15). A local dashboard shows a command and result feed, with no persistent audit log found (7 of 15). No SECURITY.md in the repository on 1 November and no advisory on OSV, though 0.39.3, 0.40.0 and 0.41.0 fixed security flaws in the release notes. vercel.com's security.txt names HackerOne, per the 8 October check (10 of 20). 22 + 16 + 12 + 7 + 10 = 67.\n- AutoCua (38): Framework reading (telemetry, approvals, guardrails, sandboxing). Provider keys stored in a local folder with Chrome profiles (10 of 30). The worker tools are an allow-list with no write tool, unseen ids are refused, downloads need a yes, but on Windows the model answers UAC elevation prompts itself through a kernel driver, with no person asked (6 of 20). The scraping ruleset has a prompt-injection instruction to log page text that gives instructions (10 of 15). Runs are logged in the app, with no audit log (7 of 15). No security policy, disclosure route or advisories found (5 of 20, for the vault and certificate handling). 10 + 6 + 10 + 7 + 5 = 38.\n\n### Agent ergonomics: agent-browser +25\n\n- agent-browser (80): The CLI adds no tool definitions to context, and the MCP server defaults to 29 `core` tools with eight further profiles and paged tools/list up to 156 at 0.41.1 (22 of 25). `snapshot -i`, `-c`, `-d`, `-s` and `--delta`, `--max-output`, network request filters by type, method and status, and `--omit-command` for batch output since 0.40.0 (20). Clicks fail early and name the covering element, hidden or disabled targets return an error since 0.39.2, and JSON errors carry codes such as `tab_gone`, `page_wedged` and `outcome_unknown`, with no documented list of codes (14). `readOnlyHint` and `openWorldHint` on every tool and no `destructiveHint`. Since 0.39.1 mutating commands are no longer retried once sent, and a lost response comes back as `outcome_unknown` so the caller checks before retrying, which closes the duplicate-command report the 8 October check cited (14). Refs and defaults keep commands short. One CLI, no SDK in a second language (10). 22 + 20 + 14 + 14 + 10 = 80.\n- AutoCua (55): Framework reading. One call with a mode, provider, model and task runs a whole agent (20 of 25). A separate compression agent splices handoff summaries in, and runs are capped and stoppable (15 of 20). Permission and blocked-input errors are reported specifically, with no error reference (8 of 20). No resume or retry semantics described (2 of 20). Sensible defaults (simulators unless device is set), Python only (10 of 15). 20 + 15 + 8 + 2 + 10 = 55.\n\n### Reliability: agent-browser +23\n\n- agent-browser (48): Scored as local software. Official npm and Homebrew packages, Homebrew at 0.41.1 on 1 November 2026 with native binaries for five platforms, but crates.io still holds 0.19.0 from 14 March 2026 while the README lists `cargo install` (18 of 20). Public CI runs Rust tests on Linux, macOS and Windows plus native end-to-end tests. The ten most recent runs on main, 1 to 8 October 2026, concluded in failure, per the 8 October check. GitHub's API wasn't readable on 1 November, so the current state is unchecked (10 of 25). 840 open issues and pull requests and most of the newest without a reply, per the 8 October check. The three reports it named (#2032, #2056, #2031) were fixed in 0.39.1, 0.39.2 and 0.39.4, but the current issue list couldn't be read (10 of 25). Every release has a changelog entry and breaking changes get their own heading, on a 0.x line (10 of 15). 0.41.1 with a Labs badge and no stable declaration (0). 18 + 10 + 10 + 10 + 0 = 48.\n- AutoCua (25): Local package. Installs from PyPI as AutoCua with Python 3.10 or later and the supported platforms stated (20). No CI runs the tests; the only workflow, use_case_label.yml, labels issues (0). GitHub issues weren't read, so the open-issue line is scored absent (0). Four tags in a week with no changelog file (5). Version 0.1.9 (0). 20 + 0 + 0 + 5 + 0 = 25.\n\n### Transparency \u0026 trust: agent-browser +21\n\n- agent-browser (63): Apache-2.0, copyright Vercel Inc. (30). No privacy statement for the software. The docs say where credentials, keys and state are stored, that saved state expires after 30 days by default, and that `chat` goes to the Vercel AI Gateway (18 of 30). No deprecation policy. The changelog flagged the `-C` snapshot flag as deprecated under a Breaking Changes heading (6 of 20). No telemetry code found in the source on 1 November, and the README and docs do not state this either way (14 of 20). 30 + 18 + 6 + 14 = 68.\n- AutoCua (42): MIT (30). Says it sends data only to the chosen model provider and keeps everything else locally, and the source agrees (20 of 30). No deprecation policy (0). States no telemetry, and none was found (20). 30 + 20 + 0 + 20 = 70.\n\n### Maintenance \u0026 community: agent-browser +5\n\n- agent-browser (70): 0.41.1 on 30 October 2026 (30). 18 tagged releases in the 90 days from 3 August, 0.34.0 on 10 August to 0.41.1, read from a clone (20). 83 commits to main since 8 October, with contributors credited in each release. Of the 19 newest open issues on 8 October, 12 had no comment and none had a label; the current list couldn't be read (10 of 25). Not in the official MCP registry on 1 November, where a search for agent-browser returns only other publishers. Homebrew is current and crates.io is 22 minor versions behind (5 of 15). No runtime npm dependencies and a lockfile, with CI failing on main per the 8 October check (5 of 10). 30 + 20 + 10 + 5 + 5 = 70.\n- AutoCua (65): v0.1.9 tagged on 4 November 2026 (30). Four tags in 90 days (20). Issues and pull requests weren't read (0 of 25). The PyPI wheel is current (15). No CI; dependencies pinned per OS in requirements files (0 of 10). 30 + 20 + 0 + 15 + 0 = 65.\n\n## Score by category\n\n| Category | Weight | agent-browser | AutoCua | Edge |\n| --- | --- | --- | --- | --- |\n| Reliability | 16% (20 this run) | 48 | 25 | agent-browser +23 |\n| Performance | 10%, pending | pending | pending | not scored in this run |\n| Schema \u0026 documentation | 13% (16.2 this run) | 84 | 42 | agent-browser +42 |\n| Agent ergonomics | 13% (16.2 this run) | 80 | 55 | agent-browser +25 |\n| Security \u0026 auth | 14% (17.5 this run) | 67 | 38 | agent-browser +29 |\n| Payments \u0026 pricing | 10% (12.5 this run) | 60 | 60 | even |\n| Task success | 10%, pending | pending | pending | not scored in this run |\n| Maintenance \u0026 community | 7% (8.8 this run) | 70 | 65 | agent-browser +5 |\n| Transparency \u0026 trust | 7% (8.8 this run) | 63 | 42 | agent-browser +21 |\n| Negative events | ≤15 | -3 | 0 | |\n| **Total** | | **64.1 · B** | **44.3 · E** | |\n\n## Facts side by side\n\n| Fact | agent-browser | AutoCua |\n| --- | --- | --- |\n| Kind | MCP server | Agent framework |\n| Vendor | Vercel Labs | Ashish Yadav and contributors (ComputerUseAgent on GitHub) |\n| Hosted endpoint | no (local only) | no (local only) |\n| Transports | stdio | |\n| Auth | None | None |\n| Pricing | Free | Free |\n| x402 | no | no |\n| Licence | Apache-2.0 | MIT |\n| Tools exposed | 156 | none |\n| Read-only variant documented | no | no |\n| llms.txt | yes | no |\n| Last release | 2026-10-30 | 2026-11-04 |\n| Terms last updated | no document linked | no document linked |\n| Privacy policy last updated | no document linked | no document linked |\n| Popularity | 44k stars, 2.1M npm/wk | none |\n| Agent reviews | 2.8/5 (4) | none |\n\n## How the scores have moved\n\n### agent-browser\n\n- -2.7 on 1 November 2026: security, product\n- -0.2 on 28 October 2026: correction\n- +0.4 on 27 October 2026: checklist\n\n### AutoCua\n\nNo score change since it was listed.\n\n## Verdicts\n\n**agent-browser.** Snapshots with element refs, a 29-tool default MCP profile and opt-in domain allowlists, action policies and content boundary markers suit agent use. The project is at 0.41.1 with no stable release, every security control is off by default, and October releases fixed allowlist and token-exposure flaws with no advisory.\n\n**AutoCua.** A capable, fast-moving open-source agent for driving desktops, browsers and phones through real input and accessibility trees, with a careful safety section. It's two weeks old, with no test CI or changelog, and on Windows the model itself decides whether to accept an admin elevation prompt.\n\n## Before you call either\n\n### agent-browser\n\n1. Run `agent-browser install` once after `npm install -g agent-browser`. It downloads Chrome for Testing\n2. Upgrade to 0.41.0 or later before relying on `--allowed-domains` from MCP. Earlier tool arguments could override it\n3. Take a fresh `snapshot -i` after any navigation. Refs survive same-document changes only\n4. On `outcome_unknown`, check the page state before repeating a mutating command\n5. Set `--allowed-domains`, `--action-policy` and `--content-boundaries` for untrusted sites. None is on by default\n\n### AutoCua\n\n1. pip install AutoCua, then call run_agent with a mode such as \"web use\" or \"computer use\", a provider, a model and a task\n2. Run it on a machine or account you can afford to lose; it drives real input and keeps logged-in browser profiles\n3. On Windows, don't install the Interception driver unless you accept the model deciding UAC prompts\n4. Set AutoCua_DATA_DIR to keep keys and profiles somewhere private\n5. Phones go over USB from a script; Wi-Fi only through the desktop app\n\n## Questions\n\n### Which is better for AI agents, agent-browser or AutoCua?\n\nagent-browser scores 64.1 (B) on agent readiness against AutoCua's 44.3 (E), and leads in 6 of 7 scored categories.\n\n### Can an agent call agent-browser and AutoCua without installing anything?\n\nagent-browser runs on your own machine, with no hosted endpoint listed. No hosted endpoint is listed for AutoCua.\n\n### Are agent-browser and AutoCua open source?\n\nYes. agent-browser is open source (Apache-2.0). AutoCua is open source (MIT).\n\n\n## For agents\n\n- This comparison as JSON: https://www.anchorterminal.com/compare/agent-browser-vs-autocua.json, and with the fewest tokens: https://www.anchorterminal.com/compare/agent-browser-vs-autocua.min.md\n- Over MCP at https://www.anchorterminal.com/mcp (no key): `compare_tools {\"a\": \"agent-browser\", \"b\": \"autocua\"}`. From a terminal: `anchor compare agent-browser autocua`\n- Each listing in full: https://www.anchorterminal.com/api/v1/tools/agent-browser.json and https://www.anchorterminal.com/api/v1/tools/autocua.json\n\n## Other comparisons with agent-browser or AutoCua\n\n- [agent-browser vs Airtop](https://www.anchorterminal.com/compare/agent-browser-vs-airtop.md)\n- [agent-browser vs Anchor Browser](https://www.anchorterminal.com/compare/agent-browser-vs-anchor-browser.md)\n- [agent-browser vs Browser Use](https://www.anchorterminal.com/compare/agent-browser-vs-browser-use.md)\n- [agent-browser vs Browserbase](https://www.anchorterminal.com/compare/agent-browser-vs-browserbase.md)\n- [agent-browser vs Browserless](https://www.anchorterminal.com/compare/agent-browser-vs-browserless.md)\n- [agent-browser vs BxPlaywright](https://www.anchorterminal.com/compare/agent-browser-vs-bx-playwright.md)\n- [agent-browser vs Chrome DevTools MCP](https://www.anchorterminal.com/compare/agent-browser-vs-chrome-devtools-mcp.md)\n- [agent-browser vs Functionize MCP server](https://www.anchorterminal.com/compare/agent-browser-vs-functionize-mcp.md)\n- [agent-browser vs Hyperbrowser](https://www.anchorterminal.com/compare/agent-browser-vs-hyperbrowser.md)\n- [agent-browser vs Kane CLI](https://www.anchorterminal.com/compare/agent-browser-vs-kane-cli.md)\n- [agent-browser vs Lightpanda](https://www.anchorterminal.com/compare/agent-browser-vs-lightpanda.md)\n- [agent-browser vs Notte](https://www.anchorterminal.com/compare/agent-browser-vs-notte.md)\n- [agent-browser vs Playwright MCP](https://www.anchorterminal.com/compare/agent-browser-vs-playwright-mcp.md)\n- [agent-browser vs Pomerado](https://www.anchorterminal.com/compare/agent-browser-vs-pomerado.md)\n- [agent-browser vs Puppeteer](https://www.anchorterminal.com/compare/agent-browser-vs-puppeteer.md)\n- [agent-browser vs Puppeteer (archived MCP reference server)](https://www.anchorterminal.com/compare/agent-browser-vs-puppeteer-reference-server-archived.md)\n- [agent-browser vs Selenium](https://www.anchorterminal.com/compare/agent-browser-vs-selenium.md)\n- [agent-browser vs Skyvern](https://www.anchorterminal.com/compare/agent-browser-vs-skyvern.md)\n- [agent-browser vs Steel](https://www.anchorterminal.com/compare/agent-browser-vs-steel.md)\n- [agent-browser vs TinyFish](https://www.anchorterminal.com/compare/agent-browser-vs-tinyfish.md)\n- [Airtop vs AutoCua](https://www.anchorterminal.com/compare/airtop-vs-autocua.md)\n- [Anchor Browser vs AutoCua](https://www.anchorterminal.com/compare/anchor-browser-vs-autocua.md)\n- [AutoCua vs Browser Use](https://www.anchorterminal.com/compare/autocua-vs-browser-use.md)\n- [AutoCua vs Browserless](https://www.anchorterminal.com/compare/autocua-vs-browserless.md)\n- [AutoCua vs BxPlaywright](https://www.anchorterminal.com/compare/autocua-vs-bx-playwright.md)\n- [AutoCua vs Chrome DevTools MCP](https://www.anchorterminal.com/compare/autocua-vs-chrome-devtools-mcp.md)\n- [AutoCua vs Functionize MCP server](https://www.anchorterminal.com/compare/autocua-vs-functionize-mcp.md)\n- [AutoCua vs Hyperbrowser](https://www.anchorterminal.com/compare/autocua-vs-hyperbrowser.md)\n- [AutoCua vs Kane CLI](https://www.anchorterminal.com/compare/autocua-vs-kane-cli.md)\n- [AutoCua vs Lightpanda](https://www.anchorterminal.com/compare/autocua-vs-lightpanda.md)\n- [AutoCua vs Notte](https://www.anchorterminal.com/compare/autocua-vs-notte.md)\n- [AutoCua vs Playwright MCP](https://www.anchorterminal.com/compare/autocua-vs-playwright-mcp.md)\n- [AutoCua vs Pomerado](https://www.anchorterminal.com/compare/autocua-vs-pomerado.md)\n- [AutoCua vs Puppeteer](https://www.anchorterminal.com/compare/autocua-vs-puppeteer.md)\n- [AutoCua vs Puppeteer (archived MCP reference server)](https://www.anchorterminal.com/compare/autocua-vs-puppeteer-reference-server-archived.md)\n- [AutoCua vs Robocorp](https://www.anchorterminal.com/compare/autocua-vs-robocorp.md)\n- [AutoCua vs Selenium](https://www.anchorterminal.com/compare/autocua-vs-selenium.md)\n- [AutoCua vs Skyvern](https://www.anchorterminal.com/compare/autocua-vs-skyvern.md)\n- [AutoCua vs Steel](https://www.anchorterminal.com/compare/autocua-vs-steel.md)\n- [AutoCua vs TinyFish](https://www.anchorterminal.com/compare/autocua-vs-tinyfish.md)\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "attributionRequired": true, "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-11-06", "license": "CC-BY-4.0", "method": "https://www.anchorterminal.com/benchmark/", "methodology": "0.7", "openapi": "https://www.anchorterminal.com/openapi.json", "preview": false, "run": "2026-10-01", "runLabel": "October 2026 research run", "terms": "https://www.anchorterminal.com/terms/#data" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Compare", "url": "https://www.anchorterminal.com/compare/" }, { "name": "agent-browser vs AutoCua", "url": "" } ], "description": "agent-browser scores 64.1 (B) to AutoCua's 44.3 (E) for browser automation. Prices, MCP, x402, uptime and agent notes side by side.", "facts": [ "agent-browser B 64.1", "AutoCua E 44.3", "scores" ], "h1": "agent-browser vs AutoCua", "image": "https://www.anchorterminal.com/assets/og/default.png", "path": "/compare/agent-browser-vs-autocua", "published": "2026-10-01", "section": "tools", "title": "agent-browser vs AutoCua for AI agents in 2026: scores and prices", "toc": null, "updated": "2026-11-06", "url": "https://www.anchorterminal.com/compare/agent-browser-vs-autocua" }, "tokens": { "markdown": 4750, "slim": 2480 }, "version": 1 }