An online converter is convenient because there is nothing to install. The document usually leaves the machine at the same time.
A privacy promise is not a technical boundary. Check the data path, the company policy and the tool’s actual network behavior.
The data path of an online converter
- You upload a file, paste a URL or sign in.
- The content travels to a remote server.
- The server parses, converts and may store it.
- The result returns to the browser.
- Retention, training use and staff access depend on the provider’s policy and systems.
Even an automatic-delete promise cannot be independently verified from the browser.
Three questions before uploading
Does the document contain sensitive information
Client names, quotations, contract terms, architecture diagrams, account screenshots and personal data require extra care.
Is external processing allowed
Many security policies forbid uploading internal documents to unapproved third-party services. Free access or encryption does not replace organizational approval.
Does it work offline
If conversion fails as soon as the network disconnects, the processing path depends on a remote service. A tool that can convert offline has a much stronger claim to local processing.
How to verify a tool
Repeat the task offline
Install the extension or open the page, disconnect from the network, and convert a small sample. Core processing should still work without a long upload wait.
Inspect network requests
Open browser developer tools, start a conversion and watch the Network panel. Look for requests that carry document text, image data or file contents to another service.
Requests for images, updates or a revocation file are not automatically document uploads. The relevant question is whether the document body is sent.
Review extension permissions
Permissions should match the job. A local conversion tool should not need unrelated access to passwords or the entire file system.
Be cautious with passwords and tokens
If a tool asks you to paste a Feishu password, cookie or API secret into a third-party page, the risk increases. Reusing a browser session is different from surrendering credentials.
Open source is not automatically safe
Source code can be inspected, but most users will not audit every line. Default configuration, permissions and actual network behavior matter more.
A store listing also does not prove that no data leaves the device. Review lowers some risk; it does not replace data-path analysis.
How Feishu Toolkit handles the boundary
Feishu Toolkit reads pages, converts Markdown, packages images, generates PDF files and stitches screenshots locally in the browser. Document content is not uploaded to our servers.
The extension contains no analytics, advertising or crash-reporting components. The public website uses analytics to measure page traffic, but the extension does not load that website script or send document content to the site.
Not every document needs the same protection
A public tutorial and a customer contract are different risk classes. A successful conversion of a public article does not justify using the same process for confidential internal material.
Classify the document, confirm policy and verify the tool. Only then decide whether to upload it.