A service that runs a startup's LinkedIn and X posting for a monthly fee should do three things: take a calendar of topics you wrote, publish from it on the days and at the hours promised, and show you every run, the failed ones included. If it also decides what to say, you are buying an agency, and the fee is paying for judgment as much as for posting. We run this routine on our own company page and our own X account with scheduled agents, 30 LinkedIn posts between September 26 and October 31 and 24 X posts between October 14 and November 1, and what follows is the month from the inside: the rules each post obeys, what the platforms charge, what came back, and what a person still had to do.
What the search returns for this question
On the morning of November 1 we put this question through Claude's search tool and kept the list: 15 results, none of them ours. Two were freelancer listings on Contra, three were pages from one agency (one titled "social media monthly posting package"), one was a "Social Starter Kit" for founders, one a scheduling tool's explainer titled "How much does it cost to have someone manage my social media?", and the other eight were pricing or help pages of posting tools, another agency's "Affordable Social Media Management" page, a Product Hunt list, a Substack post and a PR firm's FAQ. We read only the titles. Each sells a count of posts a month or explains the price of one, and none of the fifteen says what happens on the day a run fails, or offers the list of posts that actually went out. Those two things are most of what a fee for this should buy, so the rest of this post is about them.
The routine, as our two agents run it
| LinkedIn posting agent | X posting agent | |
|---|---|---|
| What it posts from | A calendar of 30 topics, each with the angle to take; it takes the next row marked pending | A calendar of measured findings, one row per post, with the form (short, thread or list) |
| How it posts | Types the post into LinkedIn's composer as the company page, in a dedicated Chrome profile kept logged in | Through the X API, with one chart drawn from the post's own numbers; it never opens a browser |
| When | Once a day, inside set hours, with a daily cap | Twice a day, at 09:20 and 13:00 US Eastern |
| The rules a post obeys | 600 to 1,300 characters, a hook as the first line, no links, no dashes, two to four hashtags, a 1080 by 1080 branded card attached | A claim on its own line, then short lines; 280 characters a segment, 300 to 1,000 for a list post; every number in the text and the chart traced to the calendar row or a report file; no client named |
| Who reads it before it goes out | Nobody; the calendar, the voice rules and the cap are the controls | Nobody; a lint reads the draft and refuses one that breaks a rule |
| Where the record is | The calendar row flips to posted with the date | A registry entry per post, with its link and the impressions read back later |
The two columns differ because the platforms differ. X permits automated original posts through its paid API and names browser automation as grounds for suspension, so the X posting agent never opens a browser. LinkedIn offers no equivalent, as that agent's notes put it, and its terms do not permit automation at all; our LinkedIn posting agent posts to our own page through the composer anyway, at a human pace with a daily cap, a risk we chose for our own page. For a client's page the same agent starts from drafts a person reviews before anything posts, and that is the setting to insist on from anyone selling this.
A service that runs a startup's LinkedIn and X posting for a monthly fee: what a month produced and what it cost
Counted from the two calendars and the X post registry on November 1.
| X | ||
|---|---|---|
| Posts | 30, from September 26 to October 31; 29 of them in October | 24, from October 14 to November 1: 16 short posts, 6 threads and 2 list posts |
| Days with no post | 4 of October's 31 days (October 1, 5, 7 and 17); 2 days carried two posts (October 6 and 18) | 5 of the 19 days: October 19 to 21, when a review hold stopped the runs, and October 24 and 25 |
| What came back | Not in our records: the weekly report reads X through the API and reads nothing from the company page's posts | The first 11 posts (October 14 to 23) had drawn between 19 and 70 impressions each when read on October 31; the 9 posts of October 24 to 30 had drawn 78 in total when read on October 30, four of them read within about a day of going out |
| What the platform charged | Nothing; the composer is free and the risk is the price | $0.015 a request, $0.20 when a post carries a URL; reading the 9 posts' impressions on October 30 cost $0.045 |
The output side is the easy part. Once the calendar is filled the posts go out, 29 of them in October on LinkedIn alone, and the four October days with none show in the calendar as gaps between dated rows. The return side is small and we publish it anyway: 78 impressions for nine X posts is what a new account got in the seven days to October 30, sixteen days after its first post, and the X agent's own notes say that a new account gets little reach for weeks and that low numbers in the first month are not a signal to post more.
The gap in the LinkedIn column is a fault in our own routine. We read what every X post drew and nothing about what the 30 LinkedIn posts drew, so on that channel we cannot yet tell a post that worked from one that did not, and a buyer should ask a provider for that column before signing anything.
On the fee itself we print no monthly figure for these agents. Each is quoted per agent by email after you order it from its page, and the prices on our pricing page ($690 a month for B2B startups, $299 for local businesses) are for the AEO service, a different thing. What we can say about cost is where the platforms' own charges sit: a dollar or two a month on X at two posts a day, and nothing on LinkedIn. Whatever a provider charges above that is paying for the calendar, the rules, the record and the person who reads it.
What the fee should buy, and what it will not
The failed runs, on the log. On October 19 our content review sent back that day's X thread because it gave one finding two sizes; the record of that hold stopped the next three scheduled runs at their first check, and the next post went out on October 22. Those three silent days are on the X posting log beside the days that worked. A provider whose log shows only successes is showing you a brochure.
A full calendar, and someone to fill it. The X run stops when its calendar runs out and writes a line a watchdog emails about; the LinkedIn calendar's 30th row was posted on October 31, and until a person adds rows that agent posts nothing. In our setup a person fills the rows ahead of time, from numbers in our own reports, and the agent never picks a topic itself. Ask who fills yours, and from what.
Not reach, and not conversations. In the X account's first ten days, October 14 to 23, two people replied to us and neither got an answer, because the posting routine had no step for it; replies belong to another agent, and a person posts those. A posting service that promises leads is promising something the posting step cannot see. What it can promise is that the posts go out on the days it said, in the rules you agreed, with a record you can read without asking, and that the record also shows the days they did not, as ours does for October 19 to 21.
How we checked
The LinkedIn counts are from the agent's topic calendar, 30 rows each with the date it was posted, and the X counts from the X topic calendar and the X post registry, which records every post that went out with its link and the impressions read back through the API; both were read on November 1. The week's figures (nine posts, 78 impressions, $0.045 in reads) are from our weekly performance record of October 30, covering October 24 to 30. The posting rules and the X API prices are from each agent's page and workflow notes; the review hold is from the content review ledger. The search results are from our Claude visibility round of November 1, one search for this question with the list kept. We did not open the pricing pages in those results, so no vendor's price is quoted here.
Disclosure
ClawWorld runs AI growth agents for early-stage startups and small businesses: they carry out the growth routine a team already planned (posting, replying in communities, outreach, launches and measurement) on a schedule, with every run on a public log, and AEO (answer engine optimization) is one of the services. The two agents in this post are among them, free to copy from their pages or priced per agent to have us run them, so we have an interest in this question. Every number above is from our own records, and the one we do not have is left blank.