Connect any agent to your workspace.
Paste one URL into Claude or ChatGPT, sign in, and ask it to draft, schedule and publish to your channels. Cursor, VS Code, Gemini CLI and anything else that speaks MCP work too.
Cancel anytime before the trial ends.
- Nothing to install or host
- Sign in, no key needed
- It only does what you allow
Works with
- Claude
- ChatGPT
- Claude Code
- Cursor
- VS Code
- Gemini CLI
- Grok
- Anything else that speaks MCP
Connecting takes three steps
The server is already running. In Claude or ChatGPT you sign in, and nothing else is needed.
- 1
Paste the URL
In Claude or ChatGPT, add a custom connector and paste the Orbitwrite URL. Once added, it is there in Claude on the web, desktop and mobile.
- 2
Sign in
Sign in to Orbitwrite, pick the workspace, and approve what it asks to do.
- 3
Ask
Ask for what you want in plain words. You can disconnect it at any time from Settings, under MCP.
https://mcp.orbitwrite.com/mcp
claude mcp add --transport http orbitwrite https://mcp.orbitwrite.com \ --header "Authorization: Bearer ORBIT_API_KEY"
URL https://mcp.orbitwrite.com Transport Streamable HTTP Auth Authorization: Bearer ORBIT_API_KEY
For scripts, CI and clients that connect with a key, create one in Settings under API keys. Settings has the snippet for Claude Code, Cursor, VS Code, Grok and the OpenAI API with the key already filled in.
Things to ask it
It checks your workspace before it changes anything, then tells you what it did.
Turn this changelog into a thread for X and one post for LinkedIn, and schedule both at the next free slot.
Your agent
- Listed your channels
- Found the next free slot
- Scheduled 2 posts
Both go out on Tuesday at 09:00. X gets a three-post thread and LinkedIn gets a single post with the full list.
What it can do
Sixty-six tools, covering the same ground as the app. Your agent sees only the ones its key allows, and every one of them is written up in the skills repo.
Channels and times
Read the channels you've connected, the posting times you keep, and which of them are still free.
Posts
Draft, schedule, move, publish, search and delete. One post can go to as many channels as you like.
Media
Upload an image or a video, attach it to a post, and see what the workspace has used this month.
Approvals
Ask a teammate to sign off, see where a request has got to, and approve or send it back.
Inbox
Read replies and mentions from every channel, answer them, and resolve, snooze or assign them. A reply that needs sign-off waits for an approver.
Tags, campaigns and voice
Group posts into campaigns, tag them, and read the brand voice the workspace writes in.
The details
An agent works under the same rules your team works under, and it has no way to skip them.
One set of rules
Whatever is calling, the checks are the same. Your permissions and each network's limits are applied on our side.
It only does what you allow
The key decides what an agent can reach. Give it reading rights and the publishing tools aren't there to pick up.
Refusals explain themselves
When we turn something down, the answer says why: a network's limit, a permission the key doesn't hold, or a post that has already moved on.
Publishing carries on in the background
Ask for an immediate post and you get an answer straight away, then a link to the live post once the network has taken it.
Images and video
Send it a file, or a link to one, and it ends up on the post. We check it against the network's rules before it goes out.
An HTTP API when there's no MCP client
Not everything speaks MCP. The same things are there over plain HTTP, same key, with docs you can try requests in.
A command line
For an agent that only has a shell, or a release script with no agent at all. Nothing to connect, nothing to install.
An SDK for your own product
A few lines and your app publishes to your users' channels. Same key, same rules, no network's API to learn.
The instructions are open source
What your agent reads about Orbitwrite is public and MIT licensed. We rebuild it whenever the product changes, so it can't quietly go stale.
One key's limits are its own
Rate limits count against the key doing the work. An agent stuck in a loop slows itself down and leaves your team working.
Your logins never leave
An agent sees your channels by name and handle. The logins behind them stay with us, and no tool has a way to ask for one.
The HTTP API is documented at orbitwrite.com/api/v1/docs. If Claude Code is what you use, the plugin page is a shorter way in; for an agent with only a shell there’s the CLI, and for your own product doing the publishing, the SDK.