Parseable

AI Agent Access

Use pb safely from AI coding agents and automation tools.

AI coding agents can use the installed pb binary through shell commands. The safest setup is to have a human create and verify the profile first, then let the agent use that saved connection.

pb login
pb status --output json

Once the profile is ready, the agent can use it directly. It does not need the API key in its prompt and should not read ~/.config/pb/config.toml.

Non-interactive profile setup

For automation, load the API key from a secret store or environment variable. The profile name is up to you.

pb profile add automation-readonly https://parseable.example.com \
  --api-key "$PARSEABLE_API_KEY" --output json
pb cloud profile add --name automation-readonly \
  --api-key "$PARSEABLE_CLOUD_API_KEY" --output json

If more than one profile exists, select the intended one before starting the agent:

pb profile default automation-readonly --output json
pb status --output json

Prefer read operations and JSON

When an agent needs to read results programmatically, prefer JSON output:

pb status --output json
pb dataset list --output json
pb dataset info <dataset> --output json
pb sql run "SELECT * FROM <dataset> LIMIT 10" --output json
pb promql run "up" --dataset <metrics-dataset> --instant --output json

Create a dedicated read-only API key or role with only the query and metadata permissions the agent needs. Never give an agent an administrator key or a shared human credential. Read-only access must be enforced by Parseable on the server; CLI instructions alone cannot prevent create, update, or delete attempts.

Keep API keys out of prompts, source control, logs, and agent instruction files.

Was this page helpful?

On this page

Non-interactive profile setup
Prefer read operations and JSON