Feature request: add `browser_downloads_list`and `browser_download_get` tools to allow discovery and retrieval of downloaded documents

[eschnou]

I'm using playwright-mcp to automate site navigation and download of documents. It works great using the current version, however the file is stored in a non-accessible temporary environment. I would like to be able to configure another download destination to be able to properly retrieve the files.

Is this already supported? If not, what would be the best way to approach this? I'd be happy to contribute a pull request, I would however appreciate some pointers.

[Skn0tt]

It's not yet supported, no. We need to be careful to not introduce a security vulnerability through this - you don't want the LLM accidentally overriding a credential file for example.

Before we add anything, let's double-check if this is necessary. Can't the LLM copy the file over from the temp directory?

[eschnou]

Hi @Skn0tt, it could definitively, however at this stage I'm not even sure the file is even stored on disk, when I try to open it in the browser it refers to a 'blob'. In the attachment, you can see the file is not on disk but in a blob.

So, you are correct that my request is maybe not the proper solution. A better approach could be tools that allow to list the download files and retrieve one, in base64, through MCP. This would also be better suited to cases where playwright is running on a remote instance, docker image, etc.

[Skn0tt]

Makes sense!

[Skn0tt]

I'll also experiment with exposing downloads as MCP resources.

[hugofreire]

Hi, is there away to access download files, i cant find them in my machine

[eschnou]

@hugofreire No. There is however a PR #162 in the work to provide tools to get a downloaded file over MCP.

[hugofreire]

@hugofreire No. There is however a PR #162 in the work to provide tools to get a downloaded file over MCP.

thanks :)