GSA Ser MCP/API

@sven

Hello everyone,

With the advancement of AI by 2026, the ways we create backlinks have also advanced significantly. While this has made life easier in some respects—and harder in others—overall, I see the net result as positive.

@sven, I’d like to make a suggestion for GSA SER today.

What do you think about implementing an MCP server?

I saw this suggested here: https://forum.gsa-online.de/discussion/34814/xrumer-23-0-8-vs-gsa-ser-full-comparison-and-a-question-for-sven-about-mcp

And an API suggested here: https://forum.gsa-online.de/discussion/34384/gsa-ser-has-an-api

The key approach is to create an MCP/API so we can connect to GSA SER and interact with it.

Through this connection, I could:

- create/edit/delete campaigns;

- view submitted/verified backlinks;

- customize campaigns—anchor text, spinning, and any other customizable settings;

We currently have excellent tools—some established, others new. For instance, SCM has evolved to include MCP support, and tools like Hermes Agent have emerged.

My current workflow is:

- A Python script performs comprehensive research and generates articles using Hermes Agent;

- The created items are imported into GSA SER.

This workflow could be fully automated—for example, Hermes could handle the research and campaign creation, importing everything directly into GSA SER without me ever needing to open the GSA SER interface.

What do you think? If you take the first step, I’m willing to contribute in some way. I’m not sure yet exactly how I can help, but could you point me in the right direction regarding what’s currently possible with SER?

Its on my list! Though there have to be other things finished first. Remember, this is a two guys company.

Sven said:

Its on my list! Though there have to be other things finished first. Remember, this is a two guys company.

You're doing great for a company with hundreds of users. Pushing updates and tools like GSA SER is a lifesaver.

Anyway, I'm also willing to donate to you for such MCP integration, and I'm pretty sure you'll gain more sales with AI evolution.

no need for a donation, but maybe someone can create a setup that can be used to test things ones I have added this.

@sven

This is what i have been thinking and if @sven add this would be great. 

For anyone willing to test this, let me know in a private message. I have done some quick test implementation on latest SER.

i would like to test it out 

I also would like to try. 

Since i run multiple copies of gsa ser on servers this would be great. 

I have enough data for now to continue the implementation. Though, Im on vacation now and can continue this work only next week. I will give detailed instructions be then.

Sven

Sven said:

I have enough data for now to continue the implementation. Though, Im on vacation now and can continue this work only next week. I will give detailed instructions be then.

Enjoy your weekend big brother

@Sven

Great to see this coming. While we wait, here's how we got a fairly hands-off setup with no API at all. It might help anyone in the same spot, and it shows where an MCP would save the most work.

Setup: two servers, 80+ projects (tier1 plus a tier2 on "use verified URLs of"). Everything runs on files, Windows Task Scheduler and plain Node scripts, and one git push delivers code to both servers. No click automation.

What works through files

- Project settings. The .prj is INI-like text, so you can patch it. But SER rewrites it from memory on save and exit, so close SER first, change only the lines you need, and diff before and after. Never regenerate a live project from scratch, or you lose everything set in the GUI since.

- New projects in a running SER. A .ser backup is a plain zip with the project files in the root (.prj, .articles...). We build one from a reference project and load it with Modify project -> Restore backup, no restart needed. (Thanks Sven for pointing to that one here on the forum.)

- Articles. Projects read articles straight from a folder via %spinfolder%. A weekly job rebuilds each campaign's pool (1000 articles, gated on uniqueness, footprint and structure) next to the live one, and swaps the folders only if every check passes. SER picks up the new pool without anyone touching the project.

- Stats. .success is the Verified column. .verify is Submitted, which is a queue, not a running total (a URL is never in both). Every week we snapshot both, diff them and send a summary to a Telegram chat: "stalled", "submitting but nothing verifies".

- Watchdog. A daily check of every .prj caught two silent killers. The notes field has to stay ASCII, because non-ASCII text gets re-encoded and doubles on every save (one project file grew to 202 MB). And skip lang list needs ISO codes, because language names get wiped and the filter quietly stops filtering.

Where files run out, which is exactly what I hope the MCP/API will cover:

- Active/Inactive isn't stored in any file, and a new project starts Active, so someone has to untick it in the GUI.

- Starting and stopping projects.

- Live Submitted/Verified per project, without parsing files.

- Changing settings or importing articles/targets in a loaded project without closing SER.

- Creating a project from a template in one call.

P.S. The day-to-day work is done by an AI coding agent over SSH using these same scripts. With an MCP it could ask SER directly instead of reading its files.

P.P.S. We're writing all of this up as a run guide and will share it when it's ready. @Sven, if you need more testers later, we're happy to try the MCP build on a 2-3-server setup.

spintax

spintax said:

Been doing very similar but still MCP helps as one stop solution,  Currenly using similar with tailscale network + hermes. Thank you for the stats clue.  I didnt note that but still been using for .prj and creating projects automatically..