What can an agent token do?
When your AI agent connects to FlurryPORT it works through a named access token, not your account password. There are two kinds, and the names in your Settings match what the agent was actually given.
AI editor (read-only)
- Read captured webhooks, with personal data masked
- Forward captures to the developer's own machine
- Check secret-setup progress by NAME only, never values
- Search and read this recipe catalog
This is the default token every agent gets. It cannot send anything to the outside world and it cannot change your account.
AI editor (read-write)
- Create capture endpoints and delivery targets
- Create and bind payload transformations
- Arm targets and post signed events to finish the pipe
It cannot:
- Read your secret values
- Delete projects, endpoints, or captured data
- Touch billing or account settings
Recipes that deliver externally need this token to finish wiring. You grant it yourself: after saving a recipe's secrets, the setup page offers [Grant write access] with this exact list, or you can generate a read-write token under Settings. Agents never request scope; the platform infers it from the recipe.
Revoking access
Both tokens live under Settings > Access tokens in your account, under the names above. Deleting or revoking one cuts the agent off immediately: every request is checked against the live token, so there is no grace period and nothing cached. Your secrets and pipes stay exactly as they are; only the agent's access ends.