My Agents
Settings › My Agents: connect Claude or your own tooling, see every tool an agent can call, and read back every call it made.
Settings → My Agents is where you let Claude, or your own tooling, work inside Ballad, and where you can see exactly what it did. It's also the first step of the setup list in your Inbox, Connect via Agent. Ballad is an MCP server, so an agent connects the same way it would to any other tool, and then works under the same review gate you do.
The page opens with your connector URL, the number of connections, how many tools they can call, and the calls made in the last seven days.
Connecting an agent
Two paths, both on the page:
- Claude: copy the connector URL into Claude as a custom connector, sign in with Ballad, and pick what it may do on a consent screen. No key is ever pasted anywhere. Step by step: Connect Claude to Ballad.
- Anything else: mint an agent key, scope it, and use it as a bearer token. Keys are shown once and can be revoked from this page.
Scopes, and what they allow
| Scope | What it permits |
|---|---|
| read | See drafts, lanes, the plan, people, signals. Always on. |
| draft | Start generation and submit drafts, which land in your review. |
| propose | Propose plan changes and launch weeks for you to decide on. |
| decide | Decline drafts, or an approved piece before it goes out; undo, set cadence. Never approve. |
| publish | Mark something published, or post it directly. |
| signals | Record metrics, events and contacts. |
The defaults are read, draft and propose. Decide and publish are off unless you turn them on.
The rules don't change because a machine is asking: an agent's drafts go through the same review as Ballad's own, an agent can't approve anything (approving is yours, or Ballad's own on lanes you set to go out on their own), and nothing reaches a channel without the publish scope.
Roles apply too. A connected app acts as the person who signed it in, so it can't do more than they can: for a User or a Viewer it's read-only for now, whatever was ticked on the consent screen. Agent keys belong to the workspace, so only owners and administrators can make or revoke them.
The tool catalogue
The page lists every tool an agent can call, grouped by what it does: reading the workspace, drafting, proposing plan changes and campaigns, recording releases and commits, managing contacts and segments, running the site audit. Each says which scope it needs, so you can see what a connection can and can't reach before you grant it.
Access and activity
Access holds your agent keys and your connected apps, each with what it may do and a revoke button that takes effect immediately.
Activity is the audit trail: every tool call, which agent made it, what it passed, whether it worked, and how long it took, next to your own decisions in the same history. Calls you make through Ask Ballad run as you, so they aren't listed here.
Next
- Connect Claude to Ballad: the five-minute setup.
- Agents & developers: the APIs behind this.
- Settings: everything else about your workspace.