Feature Request: Add Save to File for take_snapshot and evaluate_script tools

[brian-ln]

Is your feature request related to a problem? Please describe. I am unable to save off screenshots using the chrome-devtools-mcp server. Sometimes I also get too many token errors on tool results.

Describe the solution you'd like Add a save to file optional parameter to tools that return images (take_screenshot) or large amounts of data (take_snapshot, evaluate_script) .

The mcp-chrome server (and several others) supports saving images as files using the asPng optional parameter.

From the mcp-chrome server:

{
    name: TOOL_NAMES.BROWSER.SCREENSHOT,
    description:
      'Take a screenshot of the current page or a specific element(if you want to see the page, recommend to use chrome_get_web_content first)',
    inputSchema: {
      type: 'object',
      properties: {
        name: { type: 'string', description: 'Name for the screenshot, if saving as PNG' },
        selector: { type: 'string', description: 'CSS selector for element to screenshot' },
        width: { type: 'number', description: 'Width in pixels (default: 800)' },
        height: { type: 'number', description: 'Height in pixels (default: 600)' },
        storeBase64: {
          type: 'boolean',
          description:
            'return screenshot in base64 format (default: false) if you want to see the page, recommend set this to be true',
        },
        fullPage: {
          type: 'boolean',
          description: 'Store screenshot of the entire page (default: true)',
        },
        savePng: {
          type: 'boolean',
          description:
            'Save screenshot as PNG file (default: true)，if you want to see the page, recommend set this to be false, and set storeBase64 to be true',
        },
      },
      required: [],
    },
  },

Describe alternatives you've considered Claude code can't (or won't) let me save the [Image] data returned from the tool call to take_screenshot...

Additional context The mcp-chrome server (and others) also support returning images as base 64 content using the storeBase64 optional parameter.

[OrKoN]

I think it makes sense to have this. What we currently do:

1. if the image is less than 2 MB we return it as an image content according to the MCP.
2. if it is larger than 2 MB, we save it to a tmp file (automatically generated filename).

Note that the feature request is also about saving script output to file so it only partially overlaps with https://github.com/ChromeDevTools/chrome-devtools-mcp/issues/152.

[natorion]

There is still the open question if its not the MCP client's job to handle such interactions properly.

[OrKoN]

The part about saving screenshots to file was merged but saving script output to file is still to do.

[xubaifuCode]

Should we also consider adding the JS file path as an input parameter for evaluate_script? I've found that instead of having the large model temporarily write a lot of repetitive code, it's better to fix the JS file locally and let it be directly read and transferred. This can reduce token consumption.

[OrKoN]

Should we also consider adding the JS file path as an input parameter for evaluate_script? I've found that instead of having the large model temporarily write a lot of repetitive code, it's better to fix the JS file locally and let it be directly read and transferred. This can reduce token consumption.

Could you please file a separate feature request for this? and perhaps explain in more details what kind of scripts does your use case require?

[Coquinate]

I'll take a look at implementing this. Seems straightforward since screenshot already has the filePath pattern.

[sebastianbenz]

@Coquinate as suggested by @OrKoN, please create a separate issue first and list a few use cases for this first before working on this.

[Coquinate]

@sebastianbenz I believe there may be some confusion - I've worked on the original issue #153 (adding filePath parameter to save evaluate_script and take_snapshot output to file), not the separate suggestion from @xubaifuCode about JS file input.

Issue #153 already documents the use cases:

• Reducing token consumption when dealing with large script outputs or snapshots
• Handling data that exceeds MCP size limits

I've already submitted PR #248 with the implementation following the same pattern as take_screenshot's filePath parameter. Happy to discuss the approach or add more context if needed.