Skip to main content
AI assistants such as ChatGPT, Claude and Cursor can read the Telzino docs, call the Telzino API and change your AI agents for you. You describe what you want in plain language, for example “add a transfer rule for billing questions”, and the assistant reads the agent, proposes the change and applies it after you approve. This works well for:
  • Making the same change on many agents, or copying an agent for a new location.
  • Reviewing recent calls and suggesting better instructions.
  • Explaining what an agent is set up to do, in plain language.
The assistant uses the same public API as any other integration. There is no separate AI-only interface, and it can’t do anything your API key can’t.

Safety rules

Follow these rules every time you give an assistant access to your agents.

1. Use the narrowest API key

Partners create API keys. A partner creates a dedicated key for the assistant under Account Settings → API Keys and scopes it to the one reseller or organization the assistant should work on, using the key’s Access setting. A scoped key can only read and change that organization’s or reseller’s data, and it is refused for account-level endpoints such as the billing report. See Scoped API keys.
Keys are not read-only: any key that can read an agent can also change it. Limit what the assistant can reach with the key’s Access, and use the propose-then-approve flow below to control what it changes.
  • Only partners create API keys and choose their scope. Reseller admins can rotate the keys a partner issued to them, but can’t create new ones. Each key is a client ID and a client secret.
  • Give the assistant the key through an environment variable or your tool’s secret settings. Never paste it into a shared chat or commit it to a repository.
  • When you’re done, Delete the key on the API Keys page, or Rotate it if it may have been exposed. Rotation keeps the client ID and replaces the secret, and tokens issued before the rotation stop working.

2. Propose, then approve

Tell the assistant to show you the exact change before it makes it:
  1. The assistant reads the agent with Get Agent.
  2. It shows you the fields it will change, with the old and new values.
  3. You approve, and only then does it call Update Agent.
  4. It reads the agent again to confirm the change was saved.
The starter prompt builds this rule in.

3. Know how to undo

Every settings change is saved in the agent’s Change History, the same record you see in the Telzino Dashboard and the Connect App. If a change goes wrong: You can also revert from the agent’s Change History tab in either app.
Revert restores an agent’s settings. It can’t undo actions outside the agent’s settings, such as a deleted agent, a placed outbound call or a removed knowledge base source. Don’t let an assistant delete anything or place calls without asking you first.

Next steps

Runbook for AI assistants

The steps an assistant follows to change an agent safely.

Connecting tools

Give ChatGPT, Claude or Cursor the docs and a key, plus a starter prompt.

Example prompts

Ready-made prompts for common changes.
Building your own app or portal on the API instead? See the Agent Runbook for developers.