Agents and API
Your agents propose. The brand’s rules decide.
AKETRA’s API lets your tools and agents read a brand and propose work. Every action they propose goes through the same Guard as AKETRA’s own, and decisions stay with your admins.
POST /public/v1/brands/{brandId}/actions
Authorization: <API key>
{
"type": "post.publish",
"title": "Launch week teaser",
"payload": {
"integrationId": "…",
"content": "Our new model is fully autonomous."
}
}{
"observed": false,
"action": {
"status": "PROPOSED",
"autonomy": "RECOMMEND",
"holds": ["Mentions a prohibited claim: \"fully autonomous\""],
"proposedBy": "API"
}
}What happens to a proposal
- 1
Observe
No action is created. The idea arrives as an opportunity signal.
- 2
Recommend
The action waits for an admin to approve, edit or reject it in the app.
- 3
Prepare
The post is created as a draft in the calendar.
- 4
Operate
The post is scheduled within the daily limit, or drafted when the channel can’t take it as written.
What the Guard holds
- A prohibited claim in the post
- A channel that needs attention
- A channel on another brand’s app, or off the brand’s own app when it requires one
- The daily limit, once it’s reached
- Another action of the same type already running for the brand
At Prepare or Operate, a held proposal goes back to Recommend and waits for an admin, who sees the reason.
What the API can do
Read
Brands and their channels, the brand home, the Brain, goals, campaigns and their creative directions, autonomy, actions, open signals and the activity log.
Propose
Posts, Brain items, plans and campaign drafts, written by your tools or by AKETRA, creative directions for a campaign, which AKETRA proposes, and manage goals. Brain items wait for an admin to accept them, campaigns for an admin to approve them, and directions for one of the brand’s approvers or an admin to choose one.
Stays in the app
Approving, rejecting and withdrawing actions, approving, pausing, resuming and canceling campaigns, choosing and rejecting creative directions, changing autonomy, assigning channels and managing brand apps.
Get started
- Authentication
- Your organization’s API key, or an OAuth token, in the Authorization header
- Node.js
- npm install @aketra/node
- Reference
- An OpenAPI description at /public/v1/docs
- Limits
- Proposals, goal, campaign and Brain changes, requests for directions, and new posts: 90 an hour per route and organization. Reads aren’t limited
Questions
- Can an agent approve its own proposals?
- No. Approving, rejecting and withdrawing actions stay with admins in the app, and the API has no routes for them.
- Does everything an agent does go through the Guard?
- Its proposals through the brand routes do, and so do the new posts it sends through the engine’s MCP tools: they wait for an admin unless the brand allows more, or an admin turns that check off. MCP can still change the settings of a scheduled post without a check. The same key also opens the engine’s own posts routes, which schedule, change and delete posts directly, like your team’s calendar; only new posts sent through them are checked, when the brand turns that on.
- Are there webhooks?
- Not for brand events yet. The publishing engine’s webhooks report published posts; for the rest, read the brand’s activity or open signals.
- Where is the API?
- At api.aketra.co, once AKETRA Cloud opens.
More of AKETRA
Build on brands that keep their rules.
Early access opens in stages, brand by brand.