Claude Code · one connector
Publish to social from Claude Code.
One command to connect, and the release you just cut can be announced without opening anything.
Claude Code is the shortest setup of the four: a single command adds the server, the browser opens once to authorise, and the tools are available from then on. For anyone whose release process already lives in a terminal, this puts the announcement in the same place as the tag and the deploy.
The tools are the same ones every other client gets — publish, schedule, check before sending, read what landed, retry what did not. Nothing here is a reduced terminal edition, and a post scheduled from a terminal is visible and editable from any other client signed into the same account.
Publishing from Claude Code
Drawn rather than captured — but the outcomes are the ones this product really produces, including the network that could not be confirmed.
Set me up — what do you need?
Just the accounts. Connect them in the browser once and I can publish from here after that.
Done, connected
I will tell you before publishing if a network cannot take what you wrote.
Good. Drafting the launch note now
Send it over and I will check it before anything goes out.
Check what the release post would do before you send anything
dry run · four accounts
Nothing sent. This is what would happen:
That is the whole check — refusals, overflows and cost, before a single request leaves.
Where to add it in Claude Code
claude mcp add --transport http 1social <the address on the MCP page>
The address to paste is on the MCP page, printed from live configuration rather than typed into a document — so it cannot be out of date here while being correct there.
What Claude Code can do once it is connected
One command to connect
A single add command and one browser window for the authorisation. After that the tools are simply there, next to everything else the agent can already do.
A dry run that answers first
Ask what a post would do and you get refusals, overflows, missing pictures and the cost — before anything leaves, which matters most where piping into a publish is one keystroke away.
One queue, whichever way in
A post scheduled from a terminal shows up in the web app and can be edited or cancelled from any other client. Clients are doors, not separate accounts.
Answers to instructions, not flags
The tools reach you through the agent, so there is no way to pipe a file into a publish inside a shell script. For unattended work the HTTP API is the honest answer.
How it works
Connect once
~2 minLink the accounts you use in a browser. OAuth, no passwords typed into a chat, no API keys to paste anywhere.
ChatGPT
Claude
Cursor
Ask, in your own words
one sentenceTell Claude Code to publish to every network you picked, now or on a day you name. No dashboard, no tab switch.
Read the receipt
per networkOne line per network — what landed, what failed and why, and what genuinely could not be determined.
The check before publishing is the tool worth knowing about
There is a preflight tool that answers what a post will do before it does it: which networks would refuse it outright, where the caption overflows and by how much, whether a network needs an image it has not been given, and what the whole thing costs in units. In a terminal, where the temptation is to pipe something straight into a publish call, that is the difference between finding out now and finding out from a receipt. It is also the tool that makes the rest honest — the ceilings it reports are read from the same table the publisher enforces, not from a copy.
A limitation of this product. Ours to fix. From documentation, checked 2026-08-09.
When this is the wrong tool
This is not a scriptable command-line client. The tools are reachable through the agent, so they answer to instructions rather than to flags, and there is no way to pipe a file into a publish in a shell script. If you want an unattended job, our HTTP API is the honest answer rather than this.
Questions people actually ask
What is the command to connect?
A single claude mcp add with the transport and the server address, which is printed on our MCP page from live configuration. After it runs, a browser window handles the authorisation once.
Can I check what a post will do before sending it?
Yes — that is the preflight tool. It reports per-network character counts, whether an image is required, what would be refused, and the cost in units, all before anything leaves.
Are posts scheduled here visible elsewhere?
Yes. Clients are a way in, not separate accounts. Anything queued from a terminal shows up in the web app and can be edited or cancelled from any other connected client.
Something else? Ask us — a person answers.
Read next
Start where you already work.
Connect an account and post your first update in about two minutes. The first 30 deliveries each month are free.
Posts go out through each network’s official API — not scraping.