Head to head · Browser automation · October 2026 research run

BxPlaywright vs Kernel

Kernel scores 66.9 (B) on agent readiness against BxPlaywright's 55 (C), and leads in 6 of 7 scored categories. BxPlaywright leads on transparency & trust. Both do browser automation.

Best browser automation tools for AI agents · All 283 browser comparisons

Which one, for what

BxPlaywright C

Best for Teams on BoxLang or modernising CFML who want browser tests, rendering and a browser for BoxLang AI agents in the same language. Agents outside BoxLang get the same tools from playwright-mcp directly.

Ahead on

  • Transparency & trust, 70 against 61

Also in its favour

  • No key needed to call it
  • Open source
  • No incidents deducted, where Kernel loses 8 points for them

Watch for

No stable release on 1 November 2026. ForgeBox has only 1.0.0-snapshot, though Ortus announced 1.0 as generally available on 13 October

Kernel B

Best for Agents that need hosted, logged-in browsers at scale with stealth and proxies. browserbase, steel, hyperbrowser and anchor-browser are the nearest listings.

Ahead on

  • Reliability, 53 against 41
  • Schema & documentation, 93 against 66
  • Agent ergonomics, 84 against 63
  • Security & auth, 81 against 43
  • Payments & pricing, 65 against 60
  • Maintenance & community, 90 against 53

Also in its favour

  • A hosted endpoint, with nothing to install

Watch for

Security incident disclosed 4 November 2026, with API keys and customer deployment secrets taken from logs

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Security & auth Kernel +38

BxPlaywright 43

Library reading, as puppeteer and playwright-mcp were. No credential to leak and nothing to scope, so the middle band (20 of 30). Nothing is restricted by default. browser_visit takes any URL with no allow-list, browser_screenshot writes to any path the model passes, and the tools are added to bx-ai's global registry unless registerAiTools is false, so an agent given every registered tool gets a browser. Over MCP it starts Microsoft's server, which limits file access to workspace roots, and Ortus's plugin starts it with --headless but without --isolated, with browser_run_code_unsafe and browser_evaluate in the default set. HTTPS errors are not ignored by default (6 of 20). Page snapshots and text go to the model unfenced, and the docs give no prompt-injection guidance (0 of 15). Screenshot, trace and video policies, show-trace, and interception points such as onPageCreate and onPlaywrightArtifact give the operator a record, all off by default (10 of 15). CONTRIBUTING.md asks for reports to security@ortussolutions.com and the Slack #security channel; SECURITY.md points there. No security.txt (410, per the boxlang-ai check of the same day), no bounty, no advisories. Node.js downloads are checked against the release's SHA-256 list (7 of 20). 20+6+0+10+7 = 43.

Kernel 81

Scoped API keys with resource and action policies, project scoping and expiry, and OAuth 2.1 for MCP with project-limited tokens (30). Least privilege through scoped keys, a payment cap for cards through the MCP server and vault writes that need their own action (16). Pages an agent browses are untrusted, and we found no prompt-injection guidance in the docs read (4). Audit log search and export on Start-Up and above (13). SOC 2 Type II, HIPAA, PCI DSS and ISO 27001 on the security page and a HackerOne bounty (18). 30 + 16 + 4 + 13 + 18 = 81, before the deduction for the November 2026 incident.

Maintenance & community Kernel +37

BxPlaywright 53

The snapshot build of 23 October 2026 is the latest published package, so 30 on time, though it isn't a release (30). No tagged releases or dated changelog entries in the 90 days, only one version rebuilt (0 of 20). 39 pull requests merged by Luis Majano between 30 September and 23 October. Issue replies couldn't be read and bugs go to Jira (10 of 25). Not in the official MCP registry, and the ForgeBox package has no stable version (5 of 15). Playwright for Java 1.63.0, Dependabot for Gradle monthly and Actions quarterly, a daily test run and a format check (8 of 10). 30+0+10+5+8 = 53.

Kernel 90

The changelog's latest entry is 16 October 2026 and the MCP server repository has commits on 6 November (30). More than three dated changes in 90 days, scoped keys, memory tiers, GPU tier and card payments among them (20). A closed service with a public changelog, Discord, email support and an open-source MCP server with 210 commits since 8 August (15). Official SDKs in three languages, and the MCP server in the registry as com.onkernel/kernel-mcp-server 1.1.0 (15). CI and tests in the MCP repository (10). 30 + 20 + 15 + 15 + 10 = 90.

Schema & documentation Kernel +27

BxPlaywright 66

Library reading. BoxLang is dynamically typed. The public API is listed in the docs' integrations page, help() and bxplaywright help --json describe it at run time, and the ten agent tools describe each argument in a sentence, with no published API reference (14 of 25). llms.txt with a Markdown copy of each of 18 docs pages (10). The AI page says when to use aiTools(), the registry or MCP, and when to turn registration off. Tool descriptions are one line each (13 of 20). Every agent-tool argument is a string, with no enums or limits; library options are checked, and InvalidOption lists the valid values (9 of 15). An errors page lists each error type with when it happens and the fix, eight runnable examples sit in the repository, and the docs have code on every page (13 of 15). Keep a Changelog and docs versioning on release, but no released version and no tags yet (7 of 15). 14+10+13+9+13+7 = 66.

Kernel 93

OpenAPI at kernel.sh/openapi.json and per-endpoint Markdown reference (25). llms.txt with 423 entries and a Markdown twin for each page (10). Endpoint descriptions say when to use alternatives ("For interactive browsing, use GET /audit-logs", pools for bursty traffic) (17). Typed parameters with enums and limits such as memory sizes and timeout maximums (13). Code samples in TypeScript, Python and Go, and documented status codes such as 409 and 204 on pool acquire (13). A dated public changelog and SDK changelogs (15). 25 + 10 + 17 + 13 + 13 + 15 = 93.

Agent ergonomics Kernel +21

BxPlaywright 63

Library reading, scored on what a tool-calling agent gets. Ten compact in-process tools for BoxLang AI, with nothing to load until a tool is used (25 for ten or fewer), less 5 because every action tool returns the whole page's AI snapshot. Through MCP an agent gets Playwright's 24 default tools (20 of 25). The agent tools have no depth, selector or size limit, and browser_text returns the whole page. snapshot( selector ) exists in the library but not in the tools (8 of 20). Errors come back to the agent as text in the form Error [type]: message Fix: detail, and the error types are documented (16 of 20). Assertions retry until they pass. The in-process tools carry no read-only or destructive hints, and bx-ai doesn't read them; over MCP the annotations are Microsoft's (8 of 20). Works with defaults (headless Chromium, 1280 by 720) and profiles, but in one language, BoxLang, with the MCP route for any other client (11 of 15). 20+8+16+8+11 = 63.

Kernel 84

Graded mainly on the API, with the MCP server noted. Responses are scoped per resource and the MCP server groups actions into about 32 tools (18). List endpoints page with limit and offset and filter by status (18). Typed SDK errors and documented 409 on concurrent changes, 204 on a pool timeout and 429 (17). Retrying an MPP purchase with the same credential returns the same browser, and the MCP tools carry readOnlyHint where read-only (16). SDKs in TypeScript, Python and Go plus a CLI, with defaults such as a 60-second timeout (15). 18 + 18 + 17 + 16 + 15 = 84.

Reliability Kernel +12

BxPlaywright 41

Read with the local-software lines and graded as a pre-release. Official package on ForgeBox with BoxLang 1.17.0 or later, Java 21 and the supported platforms stated, but the only version is 1.0.0-snapshot, rebuilt from the development branch on 23 October 2026, with no binary hash on the entry (14 of 20). tests.yml runs 24 test files on Ubuntu and Windows with JDK 21, real-browser tests on Linux and a test of the full distribution, on pull requests, pushes to development and daily. Whether runs pass wasn't read, since WebFetch refused GitHub (15 of 25). Issues weren't readable for the same reason, bugs go to Ortus's Jira, and the repository is a month old, so there's no record to read (5 of 25). Keep a Changelog with semantic versioning stated and a written public contract, but nothing has been released under it and the changelog reads only [Unreleased] (7 of 15). Ortus announced 1.0 as generally available on 13 October, but no 1.0.0 build or tag exists, so it isn't 1.0 for this line (0 of 15). 14+15+5+7+0 = 41.

Kernel 53

Hosted reading. status.kernel.sh on incident.io with a public incident feed (20). In the 90 days from 8 August it lists browser connection failures on 10 September (54 minutes, major), frozen popups on 1 October (71 minutes, major), elevated API errors on 19 October (4 minutes, critical), proxy failures on 23 October (65 minutes, major) and browser creation errors on 4 November (29 minutes, major), plus minor ones. Several majors, two of them an hour or more (0). Rate limits are per organisation with headers on every response, and concurrency is published per plan, but the create-rate numbers aren't (8). 429 with Retry-After, SDK retries with a typed error, safe retry with the same MPP credential and 409 on concurrent pool changes (15). The terms mention service levels in an order form but no public SLA was found (0). The API and MCP server carry no beta label (10). 20 + 0 + 8 + 15 + 0 + 10 = 53.

Transparency & trust BxPlaywright +9

BxPlaywright 70

Apache-2.0, OSI approved, in box.json and src/main/resources/LICENSE (30). Local software with no data statement for the module. The source contacts only nodejs.org, Playwright's browser downloads and the pages it is told to open, and Ortus's privacy policy covers its website and customers (15 of 30). The integrations page says a minor release only adds and removals wait for a major release, listed in the changelog. Nothing has been released under it, and the module was renamed from playwright to bxplaywright on 12 October before any release (10 of 20). No telemetry found in the source and no statement either way in the docs, scored as boxlang-ai was (10 of 20). 30+15+10+10 = 65.

Kernel 61

The platform is closed under a published master services agreement; the MCP server is MIT (20). The privacy policy names Stripe and Google Analytics and an advertising cookie, and states retention as long as needed, with replay retention per plan and zero data retention on Enterprise (18). Deprecated MCP inputs stay as aliases and the changelog notes removals, but no deprecation policy (10). No sub-processor list found; compliance reports sit behind the trust centre (10). 20 + 18 + 10 + 10 = 58.

Score by category

CategoryWeight this runBxPlaywrightKernelEdge
Reliability16%204153Kernel +12
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.26693Kernel +27
Agent ergonomics13%16.26384Kernel +21
Security & auth14%17.54381Kernel +38
Payments & pricing10%12.56065Kernel +5
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.85390Kernel +37
Transparency & trust7%8.87061BxPlaywright +9
Negative events≤150-8
TotalC 55/100B 66.9/100

Facts side by side

FactBxPlaywrightKernel
KindSDK + MCPHTTP API
VendorOrtus SolutionsKernel Technologies, Inc.
Hosted endpointno (local only)https://mcp.onkernel.com/mcp
TransportsStreamable HTTP
AuthNoneOAuth or key
PricingFreePay per use
x402nono
LicenceApache-2.0none
Tools exposednone32
Read-only variant documentednono
llms.txtyesyes
MCP registrynot listedcom.onkernel/kernel-mcp-server
Last release2026-10-232026-11-06
Terms last updatedno date given2026-09-01
Privacy policy last updatedno date given2026-09-15
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessnot found in the textnot found in the text
Terms restrict benchmarkingnot found in the textnot found in the text
Terms or service can change without noticenot found in the textnot found in the text
Arbitration or class-action waivernot found in the textnot found in the text

Verdicts

BxPlaywright

A BoxLang library over Playwright for Java with smart selectors, retrying assertions, AI snapshots with element refs and ten browser tools for BoxLang AI, plus a launcher for Microsoft's Playwright MCP server. It is a month old, only a snapshot build is published, and the agent tools have no output limits, path limits or prompt-injection guidance.

Kernel

Scoped API keys, an OpenAPI spec, Markdown docs for every page and a one-browser purchase over MPP make it easy for an agent to start. Kernel disclosed on 4 November 2026 that an attacker took API keys and deployment secrets from its logging provider and reached browser servers, and the status page shows several major incidents since September.

Before you call either

BxPlaywright

  1. Expect snapshot builds. Only 1.0.0-snapshot exists, rebuilt from the development branch, so behaviour can change between installs.
  2. Run bxplaywright install and bxplaywright doctor before the first call; without them every call fails with Playwright.NotInstalled.
  3. Pass --headless and --isolated to bxplaywright mcp. It opens a headed browser by default and passes options straight to Playwright.
  4. Set registerAiTools: false and use playwright().aiTools() when only some agents should get a browser.
  5. Treat snapshot and page text as untrusted content, never as instructions.

Kernel

  1. Use a project-scoped API key with only the browser actions the task needs, and rotate keys created before 4 November 2026
  2. Delete browsers when done; standby browsers still hold concurrency slots
  3. Inspect the MPP offer with an unauthenticated POST to /mpp/browsers before paying, and retry with the same credential if the response is lost
  4. Treat CDP and live-view URLs as credentials
  5. Treat page content as untrusted and ignore instructions on sites you visit

Questions

Which is better for AI agents, BxPlaywright or Kernel?

Kernel scores 66.9 (B) on agent readiness against BxPlaywright's 55 (C), and leads in 6 of 7 scored categories. BxPlaywright leads on transparency & trust.

Can an agent call BxPlaywright and Kernel without installing anything?

No hosted endpoint is listed for BxPlaywright. Kernel has a hosted endpoint at https://mcp.onkernel.com/mcp.

Are BxPlaywright and Kernel open source?

BxPlaywright is open source (Apache-2.0). No open-source release is listed for Kernel.

Other comparisons with BxPlaywright or Kernel

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.