The VoiceMoat blog
Growth systems, the craft of writing, and how the Twitter and LinkedIn algorithms actually rank you. No virality hacks.
X publishes three official MCP servers, and the one everybody installs cannot create a post. You can prove that from its OAuth scopes, which do not include write access to posts at all. Meanwhile every guide prices this as fifteen thousandths of a dollar per post, which is the sticker price and not the bill. A realistic assistant session costs two to eighteen times that, and the difference is mostly links.
Every guide on this question says no, and every one of them is now out of date. Some time between 11 July and 5 August 2026, LinkedIn put a Develop with MCP link on its developer homepage, badged it New, and made it the first thing you see. It has documented that precisely nowhere. Here is what is verifiably there, what is still missing, and the install trap that catches people looking for an official server in the meantime.
Claude can publish to LinkedIn through a connector, using a permission LinkedIn hands out self-serve. The final item in LinkedIn's API Terms of Use is a flat ban on using the APIs to automate posting. Every guide we read cites the wrong document, and several sell scheduling while calling themselves the compliant option. Here is the clause, what Claude can and cannot actually publish, and why the connector most people recommend cannot work on claude.ai at all.
Every comparison of LinkedIn MCP servers sorts them by stars, tools and terms risk. None of them mentions that the ones using LinkedIn's sanctioned login die after 60 days and cannot renew themselves. We went through the field server by server. Only a handful use that route at all, most of those are abandoned, and the ones that never expire avoid the problem by holding a credential that lasts a year and cannot be revoked individually.
The rate card is published, and it is the least useful part of the answer. What decides your bill is that billable events are not API calls, that one attached image doubles a post's cost, that the same post costs five times more to read through one endpoint than another, and that the monthly read cap changed three times this year with no announcement. Plus the failure nobody has written down: what the API actually returns when the money runs out, which we measured because it happened to us.
The settled wisdom is that LinkedIn gives you no way to read your own post performance programmatically. That stopped being true on 8 July 2025, when LinkedIn shipped a member post analytics endpoint with eleven metrics. We read seventeen pages that rank for this question and only three name it. Here is what exists, who can actually get it, the metrics LinkedIn will not give a personal profile at all, and the storage question LinkedIn has never answered.
Every answer to this question is confident and none of them cite anything. This one quotes the clauses. LinkedIn's user agreement names browser plugins specifically, the three routes an MCP server can use to reach your account land in very different places against it, and the enforcement cases everyone repeats are about companies rather than members. Including the part most vendors leave out: our own Chrome extension sits inside the clause we are quoting, and we would rather tell you that than claim otherwise.
Is Twitter engagement down in 2026? Yes, and the corroborated benchmarks now show it: the median X engagement rate fell from about 0.035% to 0.029% to 0.015% across consecutive annual reads from RivalIQ and Sprout Social, off an already-tiny base (Buffer's 18.8M-post read puts the median non-Premium account near 0%). Real percentages where the reports give them, directional language where they don't, and the plural-cause explanation the single-cause AI-saturation narrative misses.
The creator economy has changed in seven specific ways since 2023, only three of which are getting talked about. Here is the long-horizon read on how AI has shifted the underlying structure (fluency floor, credential premium, voice premium, volume game, attention budget, hand-off economy, platform diversification) and what compounds for creators between 2026 and 2030.
How much of Twitter/X is AI-generated in 2026? No precise platform-wide percentage is verifiable, but the directional read is clear: the median post is now AI-shaped, the heavy-AI accounts are visibly distinct, and the interesting question is in which categories AI concentrates. Here is the observation-based report on AI content on X in 2026.
AI slop is the average-quality, voice-flat, fluently-incoherent content that now floods every marketing channel. It is the quiet crisis of 2026: nobody wants to name it because too many teams are producing it. Here is what AI slop actually is, why marketing teams keep shipping it, and what the alternative looks like for creators who want to keep their audience.