One bill. Ten walls between your clients.
Every client gets a workspace with its own accounts, calendar, media, members and API keys. Contractors see one client. Clients see their own calendar, read-only. You see everything, on a single invoice.
up to 10 workspaces · 50 profiles · 20 collaborators on Business
How to set up a client in four moves
One workspace per client
Its own connected accounts, calendar, media library, members, API keys and activity log. Not a folder — a boundary.
Account manager as Admin
Can invite people, connect and disconnect the client’s accounts, and publish. Cannot touch billing or delete the workspace.
Writers as Schedulers
Can draft, schedule, publish and use the AI agent. Cannot manage people or reach the connected accounts themselves.
The client as Visitor
Read-only on posts, calendar and activity. Full visibility, zero ability to move a scheduled post at 11pm.
The six things that go wrong at an agency, and what prevents each
Workspaces are isolated. A member of one has no route to another — not a hidden tab, not an API call.
Visitor is genuinely read-only, and the activity log attributes every change that does happen.
Explicit profile targeting is the default; whole-platform fan-out is opt-in. Every API key is bound to one workspace.
Account health surfaces connections whose token has expired or is close to it, before a scheduled post fails on one.
One Claude Project or custom GPT per client carries that client’s voice, and its PostMCP key reaches only that client’s workspace. Eleven clients, eleven voices.
Create a workspace, run the OAuth connect flows, paste the MCP URL into the client’s Project. Same afternoon.
Client-safe seats
Visitor is read-only in the strict sense — no edit path exists, not even a disabled one that a determined click finds.
Provenance on everything
Author on every post, activity log on every change. The AI agent appears as its own author rather than as whoever was logged in.
One invoice
Workspaces do not multiply billing. Credits are pooled on the account and spent wherever the work happens.
Agency FAQ
How do agencies keep client accounts separated?+
With workspaces. Each holds its own connected social accounts, calendar, media library, members, API keys and activity log, and membership does not cross between them. A freelancer invited into one client’s workspace cannot see or reach another client’s accounts by any route.
Can clients review and approve posts?+
Clients get a read-only Visitor seat: they see the calendar, drafts, scheduled posts and activity, and can change nothing. For AI-driven work the gate is the agent itself: create_post schedules a draft by default, Claude, ChatGPT and Cursor ask before running a tool that changes anything, and every draft sits in the queue for a human to approve, edit or delete.
How many clients can I run on one account?+
Workspaces by plan: 1 on Free, 3 on Starter, 5 on Professional and 10 on Business — with 50 connected profiles and 20 collaborators at the Business tier. All on a single bill.
Can each client have their own brand voice?+
Yes. Keep each client’s voice where the drafting happens — a Claude Project, a custom GPT or a CLAUDE.md per client — and give that agent the API key for that client’s workspace. Every tool takes a workspaceId, so the Northwind agent writes in Northwind’s voice and can only reach Northwind’s accounts. Eleven clients, eleven voices rather than one house tone.
Is there an audit trail?+
Every post carries its author, and the workspace activity log records what changed and who changed it — including the AI agent, which appears as its own author. On a shared calendar with contractors and an agent both writing, that provenance is what makes review possible.
Can I integrate it with our own systems?+
Yes. Every action is available at POST /api/tools/{name} with an API key scoped to a workspace, and multicall batches up to 20 operations per request. Reporting, intake forms and client dashboards can read and write the same queue the team uses.