Bring your own agent · Coming soon
Your agent knows your business. Soon it gets a content team.
Agent connections: Coming soonConnect the AI assistant or agent team you already use to Studio over MCP. It brings the context and the research. Studio drafts in your voice, checks the work and schedules it, and your team decides what goes out.
Already have an account? Log in
The agent that knows your customers, your market and your notes is the best researcher you have. But its output ends in a chat window, and someone still has to turn it into posts and articles, in your voice, on schedule.
How it works (once connections open)
01
Connect it.
In Studio, open Integration Hub → Agent access to find your workspace's MCP address. Point your agent at it: it opens OAuth sign-in with a consent screen, and an owner or admin approves the connection. Owners and admins are the people who can connect an agent. Open to every plan, Free included.
AG1 · STEP 16:10Product screenshot to come.
02
It reads, then suggests.
Your agent reads your workspace's topic categories and audiences, lists the ideas already on your list, adds an idea, or submits a whole content plan. A dry run shows exactly what a plan would file before it files anything.
AG3 · STEP 16:10Product screenshot to come.
03
Studio takes it from there, and your team decides.
Ideas and plans land in Studio for a person. Drafting, checks and scheduling happen in Studio. Every tool call is recorded as an agent run, and every change is logged.
AG2 · ROW 4:1Product screenshot to come.
Running agents already? Set up your voice and topics now, so Studio is ready when connections open.
What a connected agent can do
| It can | Tool | Needs |
|---|---|---|
| Read your topic categories and audiences | get_vocabulary | default |
| List ideas already on your list | list_ideas | default |
| Add an idea | create_idea | default |
| Submit a content plan (ideas, optionally briefs), with a dry run | submit_content_plan | default |
| List your campaigns | list_campaigns | campaigns turned on |
| Read a campaign's tracking | get_campaign_tracking | campaigns turned on |
| Record an outcome (always labeled as reported) | record_outcome | campaigns turned on |
| Check an article's review | get_article_review | campaigns turned on |
Limits
- Connected agents can't approve, schedule or publish. A person approves in Studio.
- Permissions are set when you sign in, and a tool without its permission isn't even listed.
- Calls are rate-limited per connection. Browser pages can't drive the connection.
- You can disconnect an agent anytime.
- If your workspace sets approval rules (Boundaries), every agent call follows them.
What's built, and what's coming
Built, opens when agent connections open
- The MCP endpoint, the eight tools above and the consent screen: built, not open yet.
- Every tool call recorded as an agent run; every change logged
- Connected agents can't approve, schedule or publish
- Disconnect anytime from Agent access
Coming soon
- Connecting your agent
- Listings in the Claude, ChatGPT and Grok directories
- A permanent public address for the connection
- Setup guides for named assistants
- Drafting and scheduling posts by agent (a person still approves every post)
- Owner-granted agent approval, with clear responsibility terms
Proof
- Each connection belongs to one workspace, approved by an owner or admin there.
- Every tool call is recorded as an agent run, and every change is logged.