agent-browser

# Fix list: agent-browser

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/agent-browser, the October 2026 research run, assessed 8 October 2026. Grade B, 66.6 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on agent-browser: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 48 out of 100, up to 10.4 more on the total

Why it scored 48: Scored as local software. Official npm and Homebrew packages at 0.38.2 with native binaries for five platforms, but crates.io 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, and the ten most recent runs on main, 1 to 8 October 2026, concluded in failure. In the newest, Windows Rust tests and the end-to-end job failed and the other jobs passed (10 of 25). 840 open issues and pull requests. Open issues from the last week include `click` reporting success on hidden elements (#2056), the CLI re-sending a command the daemon already ran (#2032) and `--allowed-domains` failing on fresh launches (#2031), most without a reply (10 of 25). Every release has a changelog entry and breaking changes get their own heading, on a 0.x line (10 of 15). 0.38.2 with a Labs badge and no stable declaration (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 2. Security & auth, 67 out of 100, up to 5.8 more on the total

Why it scored 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 (16 of 20). Content boundary markers with a per-process nonce, output caps and a written threat model with ten 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 and no published advisories. vercel.com's security.txt names HackerOne, and the changelog records hardening of the dashboard (0.35.2) and allowlist (0.32.0) (10 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 3. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: Free, self-hosted, nothing to buy, so 20 + 20 + 20 under the self-hosted rule. No payment protocol (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 4. Agent ergonomics, 78 out of 100, up to 3.6 more on the total

Why it scored 78: The CLI adds no tool definitions to context, and the MCP server defaults to 29 `core` tools with seven further profiles and paged tools/list up to 158 (22 of 25). `snapshot -i`, `-c`, `-d`, `-s` and `--delta`, `--max-output`, and network request filters by type, method and status (20). Clicks fail early and name the covering element, and JSON errors carry codes such as `tab_gone` with recovery data, with no documented list of codes (14). `readOnlyHint` and `openWorldHint` on every tool, no `destructiveHint`, and open issue #2032 reports the CLI re-sending a command after it reached the daemon (12). Refs and defaults keep commands short. One CLI, no SDK in a second language (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 5. Transparency & trust, 61 out of 100, up to 3.4 more on the total

Made of editorial 68, provenance 53.

Why it scored 61: 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, and the README and docs do not state this either way (14 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: agent-browser.dev, registered 2026-01-13 (under a year) (0 of 15)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 6. Schema & documentation, 84 out of 100, up to 2.6 more on the total

Why it scored 84: Graded on the CLI and its MCP server. Every MCP tool has a JSON Schema input with `additionalProperties` false, and `agent-browser.schema.json` covers the config file (25). llms.txt and a Markdown copy of each docs page at agent-browser.dev (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).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 7. Maintenance & community, 70 out of 100, up to 2.6 more on the total

Why it scored 70: 0.38.2 on 1 October 2026 (30). 19 tagged releases since 13 July 2026 (20). 34 commits to main since 8 September, but of the 19 newest open issues 12 had no comment and none had a label, including two off-topic ones left open (10 of 25). Not in the official MCP registry, where a search for agent-browser returned only other publishers. npm and Homebrew are current and crates.io is 19 minor versions behind (5 of 15). No runtime npm dependencies and a lockfile, with CI failing on main (5 of 10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- unchecked: why CI fails on main. We read the job list of one run (8 October), not the logs
- unchecked: the split of the 840 open items between issues and pull requests, and maintainer replies beyond the newest 40
- unchecked: whether the vercel-labs organisation has a default security policy outside this repository. The community profile lists none
- No terms of service or privacy policy governs the software, so `provenance.terms` and `provenance.privacy` are left out
- The crates.io package is at 0.19.0 and we did not confirm who publishes it
- `kind` is `mcp` because the listing shape has no value for a CLI. The CLI is the main surface and the MCP server mirrors it

## Weaknesses

- The ten most recent CI runs on main, 1 to 8 October 2026, all concluded in failure. In the newest, Windows Rust tests and native end-to-end tests failed
- Version 0.38.2 with a Vercel Labs badge and no stable release or deprecation policy
- All security controls are opt-in. By default there is no limit on navigation, actions or output
- crates.io holds 0.19.0 from 14 March 2026, although the README lists `cargo install agent-browser` as an install route
- No SECURITY.md in the repository, and 840 open issues and pull requests, with most of the newest issues unlabelled and unanswered

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Run `agent-browser install` once after `npm install -g agent-browser`. It downloads Chrome for Testing
- Take a fresh `snapshot -i` after any navigation. Refs survive same-document changes only
- Pass `--json` for machine-readable results and `--max-output` to cap page text
- Set `--allowed-domains`, `--action-policy` and `--content-boundaries` for untrusted sites. None is on by default, and the allowlist rejects `--profile`, `--cdp` and `--restore`
- Start MCP with `agent-browser mcp`, and add profiles such as `--tools core,network` only when needed

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.