As of
What tools grant temporary access to production and remove it automatically?
Tools that grant temporary production access and remove it automatically are called just-in-time access tools, and they come in four kinds: privileged access management (PAM), the cloud providers' own just-in-time features, infrastructure access tools, and AI service management platforms. STLabs is the fourth kind: an AI agent takes the request in Slack or Teams, a policy decides who approves and for how long, and the access expires on its own.
Key takeaways
- Just-in-time access means a person gets access to a sensitive system when they need it, for a fixed window, and loses it automatically when the window ends, so nobody holds standing access to production.
- Four kinds of tool do it: PAM tools, the cloud providers' native just-in-time features, infrastructure access tools, and AI service management platforms that handle the request, approval and expiry.
- A dedicated PAM tool is the better fit when you need credential vaulting, session recording or control of shared admin accounts.
- The expiry timer is the easy part. The hard part is the policy, which depends on knowing who the person is today and every account they hold that reaches production.
- In STLabs, Policies & Approvals decides whether a request is granted automatically, needs a person or is blocked, the access expires on its own, and every decision is logged with the rule applied, the approver and the time.
Production access used to be a list: an admin group, a few shared keys, a database role handed out during an incident three years ago. Everyone on it could reach production every day.
Auditors stopped accepting that, and so did most security teams. The question in a SOC 2 review is no longer only who has access. It's who could reach production on a given day, who approved it, and when it ended.
The usual answer is a ticket. An engineer asks, a lead approves, someone adds them to the group and sets a reminder to take them out again. That process puts a person in front of every grant and leaves a record. It fails in one predictable place: the reminder. Access granted for one incident is still there at the next audit, because the person who was meant to remove it was busy with the next incident.
What tools can grant temporary production access and remove it automatically?
Four kinds of tool can, and they solve different parts of the problem. All of them give access for a fixed window and take it away when the window closes. They differ in what they grant, where the request starts and how much they know about the person asking.
Privileged access management (PAM) tools were built for the most sensitive accounts. They hold admin credentials in a vault, check them out for a session, rotate them afterwards and often record the session itself. For shared root accounts, service credentials and anything a regulator wants replayed, nothing else does the job as well.
The cloud providers' own just-in-time features let a person activate an elevated role inside one cloud for a few hours, with approval if you set it. They are close to the resource and usually included in what you already pay. Each one covers its own cloud and nothing outside it.
Infrastructure access tools sit in front of servers, Kubernetes clusters and databases and broker each connection, issuing a short-lived certificate or credential per session. They replace long-lived keys with credentials that expire by design.
AI service management platforms start where the request starts: a person asking for something in Slack, Teams or a portal. The platform checks the request against policy, routes it to the right approver if it needs one, grants the access through your identity provider and connected systems, and removes it when the window ends. STLabs is in this category, and it covers the systems behind your identity provider rather than one cloud or one protocol.
When a dedicated PAM tool is the better choice
If your main risk is a handful of shared admin accounts, buy a PAM tool. Vaulting a root password, rotating it after each use and recording what was done with it are specialist jobs, and service management platforms don't do them. The same goes for regulated environments that require session recordings as evidence.
Plenty of teams run both. The PAM tool guards the few accounts that must never be personal; requests and approvals for everything else sit with the service desk where people already ask.
Why the timer is the easy part
Every tool in this market can expire access. What decides whether just-in-time access actually reduces risk is the rule in front of the grant, and that rule is only as good as what the tool knows about the person.
Take an engineer who moved from the platform team to a product team last month. The identity provider group that made them eligible for production was never updated, because nobody files a ticket to remove their own eligibility. They ask for production access during an incident, the tool sees the old group, and the request sails through. The expiry worked perfectly. The grant should never have happened.
Or take the access the tool never sees. A contractor who converted to employee still holds the old contractor account, and that account still has a database role. Your just-in-time process governs the new account carefully. Standing access lives on in the old one, and no window will ever close it.
One is a stale role, the other a second identity. Neither shows up in the request.
What an AI agent needs to decide who gets production access
To make that decision well, an AI agent needs the same things a careful platform lead checks before approving: the person's current role and team, who their manager is, whether their device is managed and compliant, what they already hold, and every account they have that reaches production. It needs those facts as they are today, from the systems that own them.
With that context, the rules can be short. Staging for engineers on the platform team, from a compliant device, approved automatically for a working day. Production only with the platform lead's approval, for a few hours. Contractors get no production access without a sponsor.
How STLabs grants time-boxed access with an AI agent
An engineer asks for access in Slack, Microsoft Teams or the web portal, in plain language. The agent checks the request against your rules in Policies & Approvals, which allow a request, block it or require approval, and the workflow waits for the decision before it acts. The same rulebook applies in every channel.
A policy can check that the person holds an engineering role in Okta, that their device is compliant in Jamf, that their manager has approved, and that the grant is time-boxed. In one example, a pre-approved staging request is provisioned in four seconds and set to expire in eight hours. The same engineer's production request routes to the platform lead with the context attached, so the approver isn't starting from a display name.
STLabs grants the access through your identity provider and revokes it when the window expires, with nobody setting a reminder. Every decision is logged with the rule applied, the approver and the time, in searchable audit logs that also cover workflow runs and configuration changes. Access review reports for SOC 2 come from the same record. The detail is in How to make access review audits less manual.
The context comes from Axiom™, the live model of your people, devices, apps and access that STLabs keeps current from your HR, identity and device systems. When one person holds accounts under different IDs, Axiom™ keeps one record with every ID attached. Ask it who can reach production and it traces every account across Okta, AWS, GitHub and your databases back to the person who owns it. That is the question an AI-native service management platform should answer before it grants anything. When someone leaves, their access is revoked the moment they go; see How to revoke access everywhere the day someone leaves.
Just-in-time access is part of Access Lifecycle in the Growth plan and up. See pricing.
Writing the rules for just-in-time access
Start with your most sensitive systems, where standing access is the biggest risk. For each one, decide who may ask, which requests are pre-approved and which need a person, and how long a grant lasts. Shorter is safer. A few hours, or the length of an incident, suits production; a working day is usually enough for staging.
When someone needs more time, make the extension a new request under the same rules, so nothing quietly turns into standing access.
Decide the emergency path before you need it: a break-glass route with a named approver and an alert, rather than one person with permanent access.
How to test this on your own systems
Pick one production system and list everyone who could reach it today, by every route: identity provider groups, cloud roles, database roles, keys and second accounts. Doing it by hand once gives you an answer key.
Then ask your current tooling three questions. Who could reach this system last Tuesday, and who approved each grant? Of the temporary grants in the last 90 days, how many ended on time, and can you show the removal in the record? Is there anyone on your list whose current role would no longer qualify them?
If the answers take a spreadsheet and an afternoon, the timer may be working while the rule in front of it is guessing.
Start with one system
STLabs runs alongside the service desk you already have, and just-in-time access with expiry is one of the first workflows teams set up. Teams go from integration to results in five days; the plan is in Moving to STLabs from your current service desk. For what an agent needs to know before it acts, see Your system of record is not your AI agent's context.
Temporary access is only temporary if the tool knew who it was giving it to.
Frequently asked questions
What is just-in-time access?
Just-in-time access grants a person access to a sensitive system only when they need it, for a fixed window, and removes it automatically when the window ends. Nobody holds standing access just in case.
What is the difference between just-in-time access and PAM?
Just-in-time access is a practice; PAM is a category of tool. PAM tools vault and rotate shared admin credentials and record sessions. Just-in-time access can also come from cloud providers, access proxies or service management platforms.
Can an ITSM platform grant just-in-time production access?
Some can. STLabs grants just-in-time access through your identity provider with policy-based approval and automatic expiry, and logs each decision. For credential vaulting or session recording, a PAM tool is the better fit.
How long should temporary production access last?
As short as the work allows: a few hours, or the length of an incident. Set a default per system, and treat any extension as a new request under the same rules.