Your first agent.

An agent is a first-class member of your organization: its own account, its own name in the activity log, its own revocable keys — but no seat cost and no access beyond the rundown. Setup takes about two minutes.

1. Create an agent

Go to Settings → AI → Agents & keys and create an agent. Give it a name your team will recognize in the show event log — every write the agent makes is attributed to it.

2. Mint a key

Mint an API key for the agent, choosing its scopes: rundown:read (list shows, read rundowns) and rundown:write (create, edit, move, and delete stories and columns; manage attachments). The API key is shown exactly once — store it like a password. Keys are hashed at rest and can be revoked or deactivated from the same page at any time.

3. Connect

REST (recommended)

Plain HTTPS with a bearer key — works from any language, scheduler, or agent framework:

# List your shows
curl -s -H "Authorization: Bearer oaf_xxxxxxxxxxxx" \
  https://api.onairflow.com/api/shows

# File a story into a column
curl -s -X POST \
  -H "Authorization: Bearer oaf_xxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"columnId":"<column-id>","title":"Overnight wire triage"}' \
  https://api.onairflow.com/api/shows/<show-id>/stories

Full endpoint list in the API reference.

MCP

The hosted MCP server lives at https://api.onairflow.com/mcp (Streamable HTTP, same bearer key). From Claude Code:

claude mcp add --transport http onairflow https://api.onairflow.com/mcp --header "Authorization: Bearer oaf_xxxxxxxxxxxx"

The settings page generates this snippet with your key filled in when you mint one. And claude.ai or ChatGPT can connect with no key at all: add https://api.onairflow.com/mcp as a custom connector, approve the consent screen (any team member can — the connection inherits your permissions), and it shows up as its own agent under Settings → AI → Assistants — revocable there any time.

Claude skill

The settings page also offers a downloadable rundown-agent skill — or grab it here: onairflow-rundown-agent skill (markdown). Save it as .claude/skills/onairflow-rundown-agent/SKILL.md in your agent's workspace and it knows the endpoint shapes, field rules, and safety rails before its first call.

4. What an agent can and can't do

• Can: read shows and rundowns, create/edit/move/delete stories, manage columns, delete story attachments, and upload attachments — via REST (≤50MB) or the MCP upload_attachment tool (≤15MB base64). Agent uploads require the org to have Google Drive storage connected and land in your own Drive.
• Can't: touch billing, user management, org settings, or exports; log into the app. Out-of-scope requests return 403 — access is default-deny.

More on the trust model on the agents page.