TRUST INFRASTRUCTURE FOR SOFTWARE + AI AGENTS

Control every action
your AI agents take.

Put policy, approval, execution boundaries, independent verification, and evidence between AI intent and real-world impact.

Install Aegis · Free See the Control Plane

Local-first where supported Explicit authorization Independent verification Signed runtime

AEGIS CONTROL PLANE DEMO WORKSPACE

Aegis Control Plane interface showing agents, actions, policies, approvals, and evidence

POLICY DECISIONdeploy.productionAPPROVAL

VERIFICATIONIndependent checksVERIFIED

AI proposes→Aegis authorizes→Software acts→Aegis verifies→Evidence remains

HARNESS

Control intent
before impact.

Aegis sits in the action path. Analysis can recommend. Authority decides whether an action is allowed, blocked, or must wait for approval.

FILEwrite · src/auth.tsALLOW

SHELLexec · destructive commandBLOCK

GITpush · protected branchAPPROVAL

DEPLOYproduction · releaseAPPROVAL

ACTION ENVELOPE

Identity. Scope. Policy. Evidence.

File, shell, git, package, secret, API, database, and deployment actions can be mediated through the same explicit trust boundary.

FILESHELLGITPACKAGESECRETAPIDBDEPLOY

CONTROL PLANE

One place to see
who can do what.

Agents, actions, approvals, policy, incidents, and evidence come together in one operating surface. The screenshot shows the Aegis interface with demo workspace data.

Aegis Control Plane showing agent identity, current actions, policies, and evidence — Product interface · demo workspace data

Know the actor.

Identity and scope stay visible before software is trusted to act.

Gate the action.

Policy decides what the current task is actually allowed to do.

Check independently.

Verification stays distinct from the system that proposed the change.

Keep the record.

Evidence explains why an action was allowed, blocked, or escalated.

DEVELOPER ENTRY

Start next to the code.
Expand when you need control.

Aegis lives in VS Code. No web signup is required to start with Community. There is no separate Aegis username and password account to create.

code --install-extension aegis-security.aegis-securityInstall details →

Real Aegis Attack Graph product capture inside the developer workflow — Real Aegis product capture · Attack Graph / Data Sentinel

Aegis Snapshot

Drop one file.
See the review surface.

A browser-local preview for the first thirty seconds of evaluation. Source stays in this tab. Snapshot surfaces deterministic review signals; it does not issue a security verdict or replace full Aegis verification.

Browser-localNo uploadNo executionPreview only

PASTE CODE OR DROP ONE SOURCE FILE LOCAL

Drop one source file here or click to choose · max 256 KB · archives are not accepted

Paste source code

PREVIEW / WAITING

Nothing is called safe before evidence exists.

Add one source file. Snapshot will map a small set of deterministic review signals locally, then hand the full-repository decision back to Aegis.

INPUTOne inert text source

NETWORKNone

EXECUTIONNone

OUTPUTReview signals, not a verdict

Need the repository, attack graph, remediation, and proof? Snapshot stops before authority begins.

Trace findings in Aegis → How full Aegis works →

PRODUCT PROOF

Don’t trust “done.”
Verify what actually held.

Aegis keeps the finding, proposed change, executed checks, and final decision separate. The result is a chain you can inspect instead of a badge you have to believe.

01 · MAP

See the path before the verdict.

Attack Graph / Data Sentinel materializes source-to-sink paths, trust crossings, evidence, and sensitive flows before Aegis asks you to trust a conclusion.

source → sinktrust crossingsevidence

Aegis Attack Graph and Data Sentinel interface — Real Aegis product capture · Attack Graph / Data Sentinel

02 · FIX + PROVE

Fixing code is not proof.

A generated patch is a proposal. Aegis keeps remediation review separate from target re-verification, regression checks, dynamic replay, rollback state, and the final decision. NOT RUN never becomes PASSED.

bounded changere-verificationregression

Aegis Secure Fix preview interface — Real product capture · Secure Fix

Aegis Fix and Prove verification interface — Real product capture · Fix & Prove

03 · DECIDE

Incomplete evidence stays incomplete.

Trusted Analysis preserves executed tasks, policy, audit evidence, and integrity state without substituting planned work for executed proof.

Aegis Trusted Analysis interface — Real product capture · Trusted Analysis

THREAT MODELRepository context Aegis Threat Model interface

SECURITY TASK PLANAuthorized execution plan Aegis Security Task Plan interface

ONE TRUST SYSTEM

From a single file
to company-wide control.

The same separation between analysis, authority, verification, and evidence scales from the developer surface to managed runtime and organization-level operations.

DEVELOPER

Team$49 / active contributorPlan a team rollout → EnterpriseCustomScope enterprise deployment →

Start here

One product.
Two obvious ways in.

Use Community to get into the product. Use Founding Pro when you want the paid verification and remediation path.

COMMUNITY$0

Install. Open a project. Inspect the evidence.

01Install Aegis from the VS Code Marketplace.
02Open a project you are authorized to inspect.
03Run Aegis and review the security record.

Install Aegis · Free →

FOUNDING PRO$19 / month

Identity. Checkout. Server-side activation. Back to Aegis.

01Start Founding Pro and establish your GitHub identity.
02Complete the secure Paddle checkout.
03Aegis waits for authoritative server entitlement before showing paid access.
04Open Aegis in VS Code with the entitled identity.

Start Founding Pro · $19/mo →

Control every action
your AI agents take.