Superagent

You can let agents work with Superagent, a security platform: review security findings, start red-team reports, and score content and packages before an agent trusts them. It suits tasks like "triage this week's findings" or "check this package before we add it".

Note: Superagent appears in the catalog on every instance; no experimental setting is needed.

Before you connect

  • A Superagent organization API key. In Superagent, open Settings → API keys and create a key just for ThinkingMach. Keys start with sk_live_.

A separate key matters here. Superagent keys are not scoped: one key reaches everything in its organization, including permanently deleting findings and starting work that uses your organization's credits. A key used only by ThinkingMach can be revoked without breaking anything else.

Connect Superagent

  1. Open Connectors and select Superagent.
  2. Read the access line above the main button. It says who the connection belongs to and which agents can use it. Select Change to narrow the agents.
  3. Paste the key into the Your Superagent key field and select Connect.
  4. Open Permissions and review the action list before any agent runs unattended.

API key is the only method. ThinkingMach stores the key as a secret and sends it to Superagent on each call; agents never see it.

Choose access

Every action starts as Allowed, so this step matters more than usual. Superagent exposes a large tool list, and some of it costs money or cannot be undone:

  • Billable work. Starting reports and triaging findings consume organization credits.
  • Permanent deletes. Deleting a finding cannot be reversed.
  • Changes to agent protection. Some actions change rules or revoke credentials for agents Superagent protects.

ThinkingMach classifies Superagent's actions conservatively to help. Only actions that list or get something, or that Superagent marks read-only, count as reads. Anything that deletes, removes, or revokes counts as destructive. Everything else counts as a write, even when the name sounds harmless — triage_finding and restore_agent_builtin_rule are writes because they start billable work or change policy.

That makes the safe setup quick: set writes and destructive actions to Ask first (or Off) and leave reads Allowed. See Set action permissions.

Try it

Start with a read you can check in Superagent:

Using Superagent, list our five most recent open security findings with their severity. Do not start any reports or triage.

Compare the list with the findings view in Superagent.

Note: Illustrative task, not a recorded test result.

Troubleshooting and limitations

Problem Likely cause Fix
The key is rejected The key was revoked or copied incompletely Create a new key under Settings → API keys and reconnect
Credits drop faster than expected Report or triage actions are Allowed Set them to Ask first on Permissions
An agent is waiting for approval The action is on Ask first See Answer a connector review request
Calls fail during a busy run Superagent's rate limits Wait and retry, or spread the work out
Needs attention The key was revoked Create a new key and select Reconnect

Limitations: one Superagent organization per connection. There is no browser sign-in, and ThinkingMach cannot narrow what an issued key can reach — use action permissions for that.

To disconnect fully, revoke the key in Superagent, then remove the connection in ThinkingMach.

Sources