Choose how you build and maintain browser work.
Playwright MCP calls the tools a site publishes. With Customaise you add your own, to any site, each with its own approval.
By Customaise · Reviewed
Compares Playwright’s library, CLI and MCP separately with Customaise 1.3.9 / MCP 3.2.8. No matched latency, token or reliability benchmark is claimed.
Library: Saved tools in Customaise
Install page controls and agent-callable actions together, with per-script settings.
Framework: Playwright’s automation surface
Use reusable code, tests, fixtures and browser contexts in a project.
Any site: Tools where none exist
AgentScripts add WebMCP tools to sites that publish none; Playwright MCP calls only what a page registers.
Side-by-side comparison
| Capability | Customaise | Playwright |
|---|---|---|
| Where it runs [7] [1] [2] | Scripts run in a connected supported desktop Chrome session. | Local or remote browsers; headed or headless operation depends on configuration. |
| Existing sign-in [7] [1] [3] | Use tabs in the browser profile connected to Customaise. Sites can still require a new login. | MCP can attach to existing Chrome or Edge through its extension; persistent profiles and stored auth are other choices. |
| Reusable code [7] [2] [4] | Installed scripts expose named actions with explicit inputs. | Functions, modules, fixtures and parameterised scripts can be reused across workflows. |
| Agent interface [7] [1] [4] | MCP and CLI expose the installed script workflow and registered page tools. | Library, CLI and MCP are distinct interfaces; choose the one that suits the agent and project. |
| Context sent to the agent [7] [1] [4] | Scripts choose what to return; an agent can also request page context. Result size is not fixed. | MCP snapshot behaviour is configurable, including a no-snapshot mode; CLI output depends on commands. |
| Browser coverage [7] [2] [5] | Focused on supported desktop Chrome extension workflows. | The library supports Chromium, Firefox and WebKit. CDP attachment is a Chromium-specific path. |
| DevTools access [11] [5] [6] | Opt-in CM_devtools uses the protocol surface exposed by chrome.debugger. | Library and CDP APIs provide browser automation; available features depend on browser and connection. |
| Chrome restricts debugger domains and targets. Customaise does not promise every CDP method or identical coverage to Playwright. | ||
| WebMCP [7] [1] [12] | AgentScripts add WebMCP tools to any site you choose, and Customaise calls them with per-tool approval. | Playwright MCP 0.0.82 and later list the tools a page registers itself, as webmcp_<tool>, when the browser has WebMCP enabled. Experimental. It does not add tools to a site. |
| Tests and maintenance [7] [2] | An agent can test installed scripts against the current page. Keep selectors and outcomes under review. | A test runner and automation APIs suit project-based checks, fixtures and CI. |
| Permissions [7] [1] [2] | Declared tools and selected approval modes affect calls; review script code too. A script an agent creates over MCP or the CLI cannot grant its own tools Allow, and an agent cannot switch on Full access. | Browser and operating-system access combine with the agent host’s execution permissions. |
| Focused AI decisions [8] [2] [13] | Free on-device AI with CM_promptAI (Gemini Nano, no key, Chrome 148 or later). Power User adds Script Decisions with Jev, OpenAI Decisions API or Gemini Decisions API. | Applications can call model APIs from code; provider setup and validation belong to that application. |
| Cost [9] [1] [4] | Free bridge allowances and Power User features; external agent and decision-provider charges are separate. | Costs depend on the agent/model, runtime and any hosted infrastructure you choose. |
What is the real difference?
A Playwright program can accept arguments, branch, expose functions and serve several workflows. Its CLI and MCP server give agents different interfaces to that automation. Customaise goes further inside the browser: AgentScripts add named tools to sites that publish none, each with its own Allow, Prompt or Deny, and the same library gives people page controls.
Customaise packages browser work as installed UserScripts and AgentScripts. The extension manages enablement, URLs, tool permissions and sharing while your chosen agent writes and plans the work.
Measure the whole workflow
A saved action can reduce repeated page inspection, but the gain depends on task, result size, model, permissions and recovery. Include setup and maintenance when comparing cost. There is no universal per-task token count for either product. [1] [2] [4]
Permissions still matter
Declared tools use Allow, Prompt or Deny. The selected approval mode and saved decisions affect whether a Prompt tool asks again; Full access permits eligible calls. Review source and permissions: tool-call controls do not isolate arbitrary script code or an agent with access to the whole computer. [7]
AI and data routing
Scripts execute in the browser. Your external agent receives the context and results you give it. Optional cloud AI, Script Decisions, remote approvals and sync have separate data flows; using the local bridge does not make the whole product offline. [10] [8]
AI in your scripts
CM_promptAI runs Gemini Nano on the device, free and with no key, in Chrome 148 or later. Power User adds Script Decisions: focused classification or scoring with Jev, OpenAI Decisions API or Gemini Decisions API, declared in the script with @grant CM_decisions and run with a saved provider key on an authorised device. Provider charges are separate. [8] [13]
When should you use which?
Customaise
You want a library of page controls and named tools that people enable, inspect and reuse from the extension.
Playwright
You need browser tests, multiple browser engines, custom automation infrastructure or unattended workflows owned by a software project.
A team can use Playwright for automated tests and Customaise for browser-side tools. Validate each workflow in its own execution environment.
Sources and scope
Compares Playwright’s library, CLI and MCP separately with Customaise 1.3.9 / MCP 3.2.8. No matched latency, token or reliability benchmark is claimed.
By Customaise. Product capabilities are based on documentation and our implementation. Recommendations are workflow judgments, not a head-to-head performance or security benchmark.
- [1] Microsoft: Playwright MCP configuration
- [2] Playwright: introduction and tests
- [3] Playwright: authentication reuse
- [4] Microsoft: Playwright CLI
- [5] Playwright: browser connections
- [6] Chrome: debugger API restrictions
- [7] Customaise: AgentScripts and permissions
- [8] Customaise: Script Decisions
- [9] Customaise: plans and allowances
- [10] Customaise: privacy and data flows
- [11] Customaise: DevTools access
- [12] Playwright: WebMCP API
- [13] Customaise: on-device AI and your own models
Related comparisons
- Customaise vs Tampermonkey: Compare a userscript manager with Customaise’s integrated workflow for building, testing and reusing UserScripts and AgentScripts.
- Customaise vs Browser Use: Browser Use offers local and cloud paths, including Browser Harness. Customaise centres the workflow on an installed library of scripts and tools.
- Customaise vs Browserbase: Compare hosted browser infrastructure with a managed script library. Stagehand’s local launch and CDP attachment are part of the picture too.