DeplAI

Application

Dashboard

Services

BYOK

Account

deplaiDocumentation

Start here

Services

Account

Help

Services

Security Agent

Security Agent scans a project, shows findings, then (if you continue) proposes source fixes and can open a GitHub pull request. The stage rail is labeled **Pipeline**.

Security Agent scans a project, shows findings, then (if you continue) proposes source fixes and can open a GitHub pull request. The stage rail is labeled Pipeline.

StageWhat goes inWhat you get
ScanThe selected project and pipeline modulesCode, dependency, secret, infrastructure, API, and optional DAST findings
ResultsThe finished scanGrouped findings you can select
Agent setupYour plan and BYOK vaultThe model that will propose fixes
RemediationSelected findings + that modelProposed diffs; you may run another round
ReviewProposed diffsYour approval, or you stop
GitHub & verifyApproved diffsPull request, then a rescan

Scan

Dialog labelWhat runs
SASTStatic code analysis. Grouped by CWE under code security.
SCAInventory + CVE match. Grouped by package/CVE under supply chain.
Full ScanBoth, plus secret scanning and infrastructure / Kubernetes / CI/CD / API checks when matching files exist.

Dynamic testing (DAST) is configured inside Security Agent with an authorized public URL. It is not a generic “attack this target” control.

Severity is critical, high, medium, or low. Remediation’s major scope is critical and high.

Leave the tab open while the scan runs. When it completes, the pipeline moves to Results.

Results

Code findings share a CWE (for example CWE-79). Supply-chain findings share a CVE and package. Counts are occurrences, not separate root causes.

Example from a real scan-report fixture:

CategoryIdentifierDetail
SCACVE-2021-23337lodash 4.17.20, severity high
SASTCWE-79Cross-site scripting grouping

An SCA fix usually bumps the package toward Grype’s fix version. A SAST fix is a source patch at the reported file and line.

Agent setup

This is the model picker, not a second scan. The page heading is Configure AI Agent; the rail still says Agent setup.

Access modeWhat DeplAI uses
PlatformDeplAI-hosted keys. Free: Best fast, Best cost. Starter and above: full list including Best coding.
BYOKA key from BYOK → Credentials.
AutoYour key if one is saved; otherwise platform.

GitHub PAT (Optional) is only for pushing the fix branch on this run and is not stored persistently.

Remediation

  1. 1Filters by the scope you chose (major vs all severities).
  2. 2Groups the same root cause across files into one work item.
  3. 3Asks the model for a patch, then checks that the diff stays in the project and addresses critical/high items.
  4. 4Waits: run another round, or take this round’s fixes.

It does not write to GitHub on this stage.

Review

You approve before anything is persisted. If you reject, you can still copy the diff and apply it yourself.

GitHub & verify

For a GitHub project, DeplAI opens or updates a pull request with the approved files (GitHub App, not your login token), then re-runs Bearer, Syft, and Grype.

Local ZIP projects save diffs on the upload instead of opening a PR.

Sessions

A Security Agent run appears under Sessions. Reopen the row for logs. The live scan does not resume from that page.

Related: How it works · BYOK · Sessions

On this page

ScanResultsAgent setupRemediationReviewGitHub & verifySessions