posting agent noun

A service your AI assistant calls to publish and schedule posts to your social networks, in your voice, with your approval. Posteora is one for Claude, ChatGPT and Claude Code.

The chat ends with "scheduled for Friday 9am" instead of "now paste this into your scheduler".

posteora.com/mcp · Sign in with Posteora (OAuth 2.1) or API key · 12 tools

Your voice

Written in your voice, not the model's.

Train your Persona on 5 to 10 of your real posts. Every draft your assistant writes through voice.write uses it, and so does every draft in the Studio. Edit the voice summary any time.

voice.write · 1 AI credit · Persona trained on your posts, summary editable

Preview first

It asks before it spends.

post.preview returns character fit per network, the exact credit cost, link and bait warnings, and TikTok readiness before anything is queued. Your assistant is told to preview first and to tell you the cost. Two rules are enforced on the server: publishing requires an idempotency key, so a retrying assistant never double-posts, and it can schedule at most two X posts a day. Every tool calls the same code as the Studio composer.

post.preview free · idempotency key required on post.publish · 2 X posts per day · X post with a link: 13 credits

Your assistant shows you this before it queues anything. Worked example.

post.preview · example

Launch day. The onboarding rewrite is live: three screens, one question each, and a first post in under five minutes. Notes and numbers: yoursite.com/launch

  • XX 156 / 280
  • BlueskyBluesky 156 / 300

14 post credits at send

X: link in the post text, 13 credits. Bluesky: 1.

Fri 09:00 · charged when it sends

Approve Cancel

The Studio

When you want to drive, open the Studio.

The Studio is the posting agent with a screen: calendar, composer, approval tray and the Activity feed, everything sent, failed and queued in one list, with an email the moment a post fails. Autopilot drafts into your empty slots and never posts by itself; that rule is enforced in code and guarded by a regression test. Nothing goes out without your yes. Ramble turns five minutes of talking into posts or a thread.

Autopilot: drafts only, enforced in code, regression test · Ramble ≤ 5 min, audio discarded after transcription · Evergreen rewrites · resumable threads

Threads

Threads with a safety net

  • Native thread composer with per-segment media and a shared tone bar
  • Resumable delivery: if posting fails after segment 3, only 4 and 5 are re-sent. No duplicates, no broken threads, no charges for segments that did not go out.

threads of 2 to 25 tweets, sent as one item · resumable delivery · 1 post credit per tweet

Autopilot

Autopilot fills your gaps. You approve every word.

It scans your week for empty slots and writes drafts into your approval tray at your best times.

"It writes drafts. It never posts anything by itself."

Enforced in code, guarded by a regression test. Your own drafts never expire, and Autopilot discloses its AI-credit spend every week.

autopilot.draft · 1 AI credit per draft · awaiting your yes · never posts by itself

Ramble

Talk for five minutes. Post for a week.

  • Ramble turns up to 5 minutes of talking into posts or a full thread, in your voice
  • Built for your phone: record on a walk, or upload a voice note
  • New posts slot into your best times

Your audio is transcribed and immediately discarded. Only the text is kept.

Ramble ≤ 5 min · audio discarded after transcription · posts or one thread

Evergreen

Your winners, rewritten, never reposted

Save proven posts to your Evergreen library. Recycling makes a genuine AI rewrite, never a duplicate: the server rejects anything matching the original and refunds the credit. Recycled drafts slot into open gaps in your week.

Evergreen rewrites · duplicates rejected, credit refunded · slots into open gaps

Stats

Every insight is a button

Engagement is pulled in for a week after each post, then turned into actions: tap one and Autopilot adjusts. Best-time learns your hours, not a generic curve. The Activity feed keeps everything published, failed and queued in one list, with an email the moment a post fails.

engagement pulled for 7 days · best-time learns your hours · Activity: published, failed, queued · email on failure

Networks

Seven networks, from chat or the Studio.

Your assistant can publish or schedule to X, Bluesky, Mastodon, LinkedIn, Pinterest, TikTok and YouTube. One post credit per network, charged when it sends.

assistant: 7 networks · Studio: 7 · 1 post credit per network at send

New networks plug into the same queue, the same credits and the same tools your assistant already calls.

How it works

Start free. Your first five posts are on us.

1

Sign in with a magic link.

No password, no card. Your account exists the moment you click the email.

2

Connect a network and your assistant.

One network, then one command or one connector. Your first 5 posts and 10 AI credits are free, topped up to 5 AI credits every day.

3

Upgrade when the 5 run out.

$9.99 a month keeps it going. Posts are charged only when they actually go out.

5 post credits · 10 AI credits to start · topped up to 5 a day · Premium $9.99/mo or $99.99/yr

Your keys

Your keys. Revoke any assistant in one click.

The server acts only on your own account and sees only the arguments of each tool call, never your conversation. Tokens are scoped and expire. The Connected via MCP list in your dashboard shows every assistant and lets you revoke it; regenerating your API key signs all of them out.

Dashboard > AI Assistants > Connected via MCP · scoped, expiring tokens · API key under Account · every call in an audit log

How the connection works
Check

Pricing

Pay for posts that send, not for accounts, networks or seats.

Free to start: 5 post credits, 10 AI credits to start, topped up to 5 every day, no card, every network, and both the Studio and the assistant connection. Premium is $9.99 a month or $99.99 a year at the founding price (list $19.99 and $199.99), kept for as long as you stay subscribed, for 100 post credits and 3,000 AI credits a month, every network included. The same credits are spent whether the run came from your assistant or the Studio, and the cost appears in the preview before anything is queued: 1 post credit per network, 1 per tweet in a thread, 13 for an X post with a link, charged only when the post goes out. Top-ups exist for heavy months. No agent tier, no API plan, no per-assistant pricing.

5 post credits · 10 AI credits to start · topped up to 5 a day · Premium $9.99/mo or $99.99/yr

Premium
$19.99$9.99/mo Founding price, kept for as long as you stay subscribed $199.99$99.99/yr Founding price · 2 months free vs monthly
Start free, connect your assistant

Open the Studio instead

Cancel anytime · your price never rises while subscribed

Try free first: your first 5 posts are on us, plus 10 AI credits to start, topped up to 5 every day. No card. Both the Studio and the assistant connection. Start free
100 post credits / mo
3,000 AI credits / mo
7 networks, all from chat

Everything included

  • All seven networks at one price. Nothing is charged per connected account
  • The Studio: calendar, composer, approval tray and Activity feed
  • MCP server for Claude, ChatGPT and Claude Code: your assistant posts in your voice
  • Autopilot drafts into empty slots; you approve every word
  • Ramble: five minutes of talking becomes posts or a thread
  • Threads with per-post media and Evergreen rewrites
  • Persona: trained on your real posts, summary editable
  • Stats insights that re-tune Autopilot
  • All 56 creator tools, no account needed
  • Top-ups for heavy months
  • Every future feature included
What a run costs show
  • post.preview, fit and cost per networkfree
  • voice.write, a draft in your trained voice1 AI credit
  • post.schedule or post.publish, per network1 post credit at send
  • Each tweet in a thread1 post credit
  • X post with a link13 post credits
  • Scheduling, cancelling, the calendarfree

same credits from your assistant or the Studio · charged when the post goes out

Frequently asked questions

A posting agent is a service your AI assistant calls to publish and schedule posts to your social networks, in your voice, with your approval. A scheduler is a place you go to; a posting agent is the thing your assistant uses. Posteora is a posting agent for Claude, ChatGPT and Claude Code, served as a Model Context Protocol server at posteora.com/mcp, and it comes with a Studio for the days you want to post by hand.

It acts only on what you ask. The server tells every assistant to preview first and state the credit cost, and two rules are enforced on the server: publishing requires an idempotency key, so a retry never double-posts, and at most two X posts can be scheduled per day. Autopilot, the Studio feature that fills empty slots, only writes drafts into your approval tray; it never posts by itself, which is enforced in code and guarded by a regression test.

Claude in claude.ai and Claude Desktop (add posteora.com/mcp as a custom connector and sign in with Posteora), Claude Code (one command, then Authenticate), and ChatGPT (add it as a connector in developer mode). Clients that only take a static header can use your API key from Account. A connector brief for Meta Muse is published at /muse/brief.txt.

From the chat or the Studio: X, Bluesky, Mastodon, LinkedIn, Pinterest, TikTok and YouTube, including threads of 2 to 25 tweets sent as one item, pictures up to 5 MB and MP4 or MOV video up to 100 MB. There is no Instagram, Facebook, Threads or Reddit publishing today.

Every account starts with 5 post credits and 10 AI credits, topped up to 5 AI credits every day, with no card. Premium is $9.99 a month or $99.99 a year at the founding price, kept for as long as you stay subscribed, for 100 post credits and 3,000 AI credits a month across every network. A post costs 1 post credit per network, 1 per tweet in a thread, and 13 for an X post containing a link; voice.write costs 1 AI credit and post.preview is free. Scheduled posts are charged only when they send, and the same credits are spent whether the run came from your assistant or the Studio. Top-ups exist.

No. Every plan, free or Premium, includes all seven networks, and the price does not change with how many of them you connect. You connect one account on each network. Many schedulers price by connected account, so four accounts cost four times the base price before a single post goes out. Here the only thing that scales is the post itself: 1 post credit per network it is sent to, charged when it sends.

It sees only the arguments of each tool call, never your conversation, and acts only on your own account. Every call is kept in an audit log and everything it schedules appears in your Studio Activity like any other post. Revoke any assistant from Dashboard, AI Assistants, Connected via MCP; regenerating your API key under Account signs every assistant out.

Early subscribers pay the founding price, $9.99 a month or $99.99 a year, against a list price of $19.99 and $199.99. That rate is kept for as long as your subscription stays active, even if Premium's price goes up for new customers later. No code needed.

No. Whatever rate you subscribe at is yours for as long as your subscription stays active; price increases only ever apply to new customers. Founding subscribers keep $9.99 a month after the founding window closes. If you cancel and later resubscribe, the then-current pricing applies.

No. On Premium, post credits and AI credits reset on each billing cycle and unused credits are cleared when the next allotment arrives. On a free account the daily refill brings you back to 5 AI credits; it does not stack. Top-ups are the exception: they are added to your balance rather than resetting it.

No. Every tool at /tools/, from character counters and hashtag generators to the engagement rate calculator and the QR code generator, works without an account, no card, no signup. A handful of AI-backed tools (the reply generator, reply audit, tone rewriter, hashtag generator and similar) are capped at 3 free uses a day per browser. The rest are unlimited.

Your training posts and voice summary live in your Posteora account. They are sent to the AI provider only when a draft is written with that Persona, whether by voice.write, the Studio or the browser extension for X, and they are not used to train the provider's models. Edit the voice summary or delete any Persona at any time from /dashboard/persona.

It is transcribed and discarded straight away; only the text is kept, and nothing is stored or replayable by us. Ramble takes up to 5 minutes of talking and turns it into posts or a thread in your trained voice. It works on your phone, and you can upload a voice note instead of recording live.

From the blog

View all →