Replit Security

    Replit applications are checked before go-live and monitored afterwards.

    Before every publication the Security Agent looks for vulnerabilities, exposed data and unsafe dependencies; afterwards Auto-Protect reports new vulnerabilities. We agree with you from which finding nothing is published.

    For IT and security teams that approve their business teams' applications on Replit and want to keep watching them after go-live.

    • Security Agent
    • Security Center
    • Auto-Protect
    • Secrets
    • SBOM
    • SOC 2
    01

    Checked before it goes live

    The team wants to publish a new version of the staff portal. The scan finds a library with a known vulnerability and a place where HR data ends up in the log. Publication waits, the agent proposes the fix, we review it, and the next scan is clean. Three weeks later a flaw becomes known in another library, and Auto-Protect has the patch ready.

    02

    How we secure applications

    We combine the platform's checks with our own review. Examples:

    • A scan before every publicationStatic analysis, dependency checks and a search for sensitive data, reviewed by a person.Result: known vulnerabilities do not go live.
    • Rules for the organisationOn the Enterprise plan the scan is mandatory, and above a severity nothing is published.Result: the rule applies to every team and every publication.
    • Secrets and accessCredentials in the secrets, injected at runtime and invisible to the agent, rights per role in the workspace.Result: no key sits in the code or in the chat history.
    • Monitoring after go-liveAuto-Protect reports new vulnerabilities and prepares patches, the Security Center shows them across all applications, with a software bill of materials.Result: your IT always sees what each application contains.
    • Our reviewRoles, rights and data flows read and tested before the application is opened to everyone.Result: a person checks what the tools do not find.
    03

    What security needs first

    A person from IT or security who assesses findings and approves exceptions. A severity above which nothing is published. A decision on who receives Auto-Protect's reports. And, for the Security Center and mandatory rules, the Enterprise plan.

    04

    What you have at the end

    A clean scan before every publication, with documented exceptions. Rules that apply to all teams. A software bill of materials for every application. Tests for roles and rights. And a procedure for how new vulnerabilities are reported, assessed and fixed.

    What you receive from us

    The review report before go-live.

    Before an application goes live we check sign-in, data, secrets, dependencies and operation, and record what was fixed. Shown here as a sample with invented findings.

    Review report

    Sample

    Review report, spare parts customer portal

    Customer
    Muster AG, Bern (fictional)
    Application
    Spare parts customer portal, built by the service department with Replit Agent
    Reviewed on
    27 October 2026
    Contact at CNEXT
    Named in the real report
    • 24checks
    • 20passed on the first run
    • 4fixed before go-live
    • 0open items

    Summary

    The service department built the portal itself in three weeks. Customers see their devices in it, order spare parts and track delivery. Four items had to be fixed before go-live, among them an API key in the code. All four are fixed and re-tested.

    No.Finding and measureSeverityStatus
    P-03The ERP API key was in the code. Moved into Replit secrets, key renewed.HighFixed
    P-07Order overviews could be opened via direct links without signing in. Access now requires sign-in.HighFixed
    P-12No database backup agreed. Daily backup set up, restore tested.MediumFixed
    P-18Two packages with a known vulnerability. Upgraded to the current version.MediumFixed

    Operation

    • Owner: the service department, with IT as deputy. Changes go through the repository and a second person approves them.
    • Next review after major changes, at the latest in April 2027.

    Next steps

    With your approval the portal goes live. The report stays with the application and shows later what was checked before launch.

    Samples with invented details for a fictional Muster AG.

    Sound familiar? Tell us where you stand today.

    Contact

    What's new.

    As of: 24 October 2026

    • As of 24 October 2026

      The Security Agent checks before publication.

      The scan combines static analysis and dependency checks with a search for sensitive data and a review by a model. Auto-Protect watches published applications for new vulnerabilities and prepares patches.

      Source: Replit: Security
    • As of 24 October 2026

      Security Center and mandatory rules on the Enterprise plan.

      The Security Center shows vulnerabilities across all of the organisation's applications, notifies the owners and exports a software bill of materials. Administrators make the scan mandatory and block publication above a severity.

      Source: Replit Docs: Security
    • As of 24 October 2026

      SOC 2 Type II, ISO 27001 in progress.

      Replit cites a SOC 2 Type II audit and describes itself as GDPR compliant; according to Replit, ISO 27001 certification is in progress.

      Source: Replit: Security

    What IT security asks about Replit.

    Is Replit's scan enough?

    It finds known patterns: unsafe dependencies, typical vulnerabilities, exposed data. Whether roles work correctly and data flows only where it should, a person checks with tests. Both belong before every publication.

    Who receives Auto-Protect's reports?

    The person you name for it, and on the Enterprise plan the owners of each application through the Security Center as well. How fast each severity is fixed we agree with you.

    Is Replit ISO 27001 certified?

    Not yet; Replit lists the certification as in progress and refers to a SOC 2 Type II audit. We request the report and the data processing agreement from Replit for your review.

    Where do credentials sit?

    In the project's secrets, stored encrypted and injected at runtime. They have no place in the code, the repository or the chat with the agent; the scan raises an alert if one turns up there.

    Agree on a useful next step

    Tell us about your current workflow, the systems involved and the result you need. In our first conversation, we identify what else we need to know to plan a useful next step.