
Argent with PostgreSQL user tables
Some people only exist in an application's own user table. Argent Strategic Intelligence reads that PostgreSQL table, read-only, so those people can own AI agents too.
What Argent does with your database
Older applications often keep their own user table. Argent can read it, so the people in it are governed like everyone else.
- People and their status. Argent reads them from your user table, together with each person's state (active or inactive).
- An owner for every agent. Every agent registered with Argent must have an active human owner, and Argent checks that owner on every request. People from your database are those owners. See Identity Tethering.
- No owner, no answer. A request from an agent whose owner is missing or switched off is refused, and nothing reaches the AI model.
- Attribute mapping, reviewed by a person. Argent reads the fields your database exposes and proposes how they map onto Argent's (who is who, email, status, department). An admin reviews it, edits it and saves it as a new version; earlier versions stay on record.
How it connects
A read-only pull. Argent reads from your database on a schedule; nothing in your database is changed.
- In the Argent console, open Integrations, Identity sources and choose Add connector, type JDBC.
- Give it a display name and enter the settings below.
- Connect, then Test: Argent pulls a sample from your database to prove the link works end to end.
- Open Attribute mapping, run the mapper, review the proposal and save it. Saving the first mapping version turns the connector on.
- Check the result under Identities, People: each person shows JDBC as their source.
What you configure
| Setting | What to enter |
|---|---|
| Connection | A PostgreSQL connection string template with no password in it, plus the vault path of the password. Argent opens its own pool for your database and never reuses its platform connection. |
| Identity table | Argent reads your schema (read-only) and the mapper proposes which table holds the people and which column is the lasting key. Tables that are not people, such as orders or sessions, are rejected and listed as declined. |
Argent only reads. It runs no statement that writes to your database, and a guard stops the connection from reading Argent's own files or falling back to Argent's own database settings.
- The secret stays in your vault. Argent's configuration holds only the vault path of the credential, never the credential itself. Argent works with HashiCorp Vault and OpenBao, AWS Secrets Manager, CyberArk Conjur and Infisical.
- Guarded dial-out. Every identity connector dials out through a guarded client that refuses private, loopback, link-local and cloud-metadata addresses. Turn on the egress allowlist and only the addresses you list are reachable. A directory, database or Keycloak server on your private network is reached by listing its private address range in the deployment's tenant private-range setting; cloud-metadata addresses stay refused whatever is listed.
When someone leaves, their agents stop
Argent does not have to wait for your database to send an event. On its next check it compares who is still active with the people it holds.
- Switched off in Argent. A person who is missing from your database, or switched off there, is deactivated in Argent. The change is signed into the audit record in the same step.
- Their agents are held. The keys of every agent that person owns are held, and each request from those agents is refused before it reaches a model.
- Access removed. Their access in Argent is removed in the same transaction, so nothing they were granted lingers.
- Your hand decision sticks. A person an admin switched off by hand stays off, even if a later directory sync reports them active. Only an admin can turn them back on.
A leaver is not deleted. Their agents are refused, held and kept on the record until an admin hands them to a new owner. See Owner Offboarding.
Where it fits in Argent
- Capabilities: Identity Tethering, Owner Offboarding, Keyless Agents, One-Click Audit Evidence.
- Goals: agent identity and ownership, audit and accountability, compliance evidence.
- Use cases: securing AI coding agents, rolling out copilots and chat safely, AI in regulated and air-gapped environments.
- Start in watch mode. Argent can record what it would have done before it enforces anything. See Watch Before You Enforce.
- See it run: take the walkthrough, browse all integrations, or book a demo.
Want this running on your own use cases?
Book a demoThe vendors with their own page in this group. Open any logo for its page. SailPoint's mark opens IdentityIQ. Also see SailPoint Identity Security Cloud.