Agent Tooling
Agent Access Codes
Scoped credentials that let AI agents deploy and manage your infrastructure within defined limits.
Overview
Agent Access Codes are separate from API keys. They're designed specifically for AI agents with built-in scoping for products, permissions, and budgets.
Format: ea_live_{random_64_hex}
Example: ea_live_7f3a9b2c4d5e6f1a8b9c0d1e2f3a4b5c... Agent codes are shown once at creation. Store them securely. If lost, revoke the old code and create a new one.
Creating an Access Code
- Navigate to Account → Agent Access in the user menu
- Click Create Agent Code
- Enter a descriptive name (e.g., "Cursor Agent", "CI/CD Deploy")
- Select which products the agent can access
- Choose permissions (Deploy, Manage, Delete)
- Optionally set a budget cap with soft or hard enforcement
- Copy the generated code immediately
Scoping
Products
Control which Edge products the agent can access:
Compute
VMs, scaling
CDN
Deployments, domains
Storage
Buckets, uploads
DNS
Zones, records
Permissions
| Permission | Allows |
|---|---|
| Deploy | Create resources, upload files, deploy applications |
| Manage | View, update, and manage existing resources, check health |
| Delete | Delete resources and teardown projects |
Budget Caps
Set a monthly spending limit for the agent:
Soft Cap
The agent receives a warning when approaching the limit but can continue operating. Useful for flexibility with awareness.
Hard Cap
The agent is blocked from creating new resources once the limit is reached. Useful for strict cost control.
Authentication
Pass the agent access code as a Bearer token:
GET https://edge.network/agent Authorization: Bearer ea_live_7f3a9b2c4d5e6f1a...
Managing Codes
- Edit — Update the name, products, permissions, and budget cap at any time
- Revoke — Immediately invalidate a code. Agents using it will lose access instantly
- Last used — See when each code was last used to identify inactive codes
- Multiple codes — Create separate codes for different agents or environments
Best Practices
One code per agent
Use separate codes for each AI tool (Cursor, CI/CD, etc.) for clear audit trails
Least privilege
Only enable the products and permissions the agent actually needs
Set a budget cap
Even a generous soft cap gives you visibility into agent spending