VICTORIA, AUSTRALIA — 16 August 2026 — goldenSoftwareENGINEERS®, the technology company behind snapWONDERS, today released an open-source Model Context Protocol (MCP) server that allows AI assistants to run the company's forensic analysis, steganography and format-conversion tools on files stored on a user's own computer.
The release is accompanied by a change to the snapWONDERS API: files can now be submitted in a single HTTP request, rather than only through the chunked resumable protocol the platform has used since launch.
Why this required an upload change first
Since its earliest versions, snapWONDERS has accepted uploads exclusively over TUS, an open resumable-upload protocol that splits a file into a sequence of requests and can recover from a dropped connection. That design reflects the platform's users, who routinely submit large media files over unreliable networks — including the Tor and I2P networks, where snapWONDERS has maintained full service since 2020 and 2022 respectively.
The protocol is well suited to that task, but it requires the calling software to implement a multi-step exchange correctly before any file can be processed. For a human developer this is a known, if tedious, piece of work. For an automated caller such as an AI assistant, it is a more serious obstacle: when the exchange goes wrong, the caller generally cannot distinguish a mistake of its own from a refusal by the service.
The API now also accepts a direct upload — a single request carrying the file — for files up to 95 MB. The resumable protocol remains available and unchanged, and continues to handle larger files and unreliable connections.
What the MCP server does
Model Context Protocol is an open standard for connecting AI assistants to external tools. snapWONDERS has offered a remote MCP endpoint for some time, but a remote server cannot access files held on a user's own device, which limited its usefulness for media work.
The new server runs locally on the user's machine. It exposes four operations — forensic analysis, hiding a file inside a cover image or video, revealing hidden content, and format conversion — each of which completes a whole task. Analysis returns its findings directly to the assistant; the hide, reveal and convert operations save their output files to disk, beside the original or in a directory the user nominates.
The company noted the limitations plainly. While the server process runs locally, the analysis itself continues to run on snapWONDERS infrastructure, and files are transmitted for processing under the same handling policies as any other upload. The server also has no ability to list or search a filesystem — it acts only on paths supplied with each request — but it can read any file the user running it can read, so the AI client's own tool-approval prompt remains the effective control over what is sent.
Availability
The MCP server is published on GitHub and the Python Package Index. As with the company's existing Python, JavaScript, PHP and Go client libraries, it is released under the MIT licence — the permissive terms cover the open-source package itself, not the snapWONDERS platform or service it calls. Both the server and direct upload are usable on the free tier. The API's OpenAPI specification, which documents both upload paths, remains publicly published.
About snapWONDERS
snapWONDERS is a digital media platform developed by goldenSoftwareENGINEERS®, founded by Kenneth Springer. The platform addresses privacy, accessibility and metadata concerns in digital media — offering tools for image analysis, media conversion, and file concealment through snapWONDERS Vaultify. snapWONDERS and Vaultify are available on the clearnet, Tor and I2P.
Developers: snapwonders.com/developers Contact: press@snapwonders.com
