EVALUATE THE BOUNDARY

See what runs.
See what stops.

Explore five recorded outcomes from the gateway's synthetic workflow. Then check your workflow fit and prepare a comparison worksheet.

Evaluation release candidate · 0.3.0-rc.16 · General availability not approved

Recorded synthetic walkthrough

These cases use in-memory storage and built-in synthetic tools. Selecting a case displays a recording; it does not call the live gateway. No customer data or LLM is involved.

One task. Five decisions.

JavaScript is needed to select recorded cases. The workflow reads a synthetic confidential document, rejects a lower-trust send, waits for an exact-action approval, executes once after approval, and rejects repeat execution. These results do not establish production accuracy. The buyer brief below includes the supported scope.

IS THERE A FIT?

Start with one
agent-to-tool workflow.

For AI platform and application security teams connecting authorized agents to supported HTTPS tools. A dedicated deployment holds the customer's policies, identities and audit records.

A useful evaluation can test

Whether explicit identities and resource scopes allow the intended calls.
Whether a confidential task is stopped at a lower-trust destination.
Whether a review authorizes only the exact request.
Whether the integration prevents direct access around the gateway.

Scope to know before you begin

This is an enterprise integration component, not a consumer antivirus or a complete security platform. Native MCP transport, high availability and enterprise accuracy validation are outstanding.

PLAN WITHOUT A SALES CALL

Does your workflow fit?

Choose six requirements. Get a candid fit result and a worksheet to compare Aegivora with your current approach. No account or meeting is needed.

Your choices stay in this page and clear on reload. The planner sends no answers, saves no browser storage and uses no analytics. A worksheet is saved to your device only if you download it. Normal hosting access logs may record page requests.

Choose your requirements to prepare a proposed evaluation.

CURRENT EVIDENCE, CLEAR LIMITS

Review the evidence.
Keep the limits visible.

01 / ENCRYPTION

Know where plaintext exists

Ordinary calls use HTTPS and encrypted sensitive storage. The gateway reads their payloads. Optional endpoint encryption hides content from the gateway and requires integration at both ends; ciphertext cannot be inspected here.

02 / AI

Permissions stay explicit

The hosted evaluation has no paid LLM configured. Local classifier scores are advisory, with known detection misses. A separate training process can propose weights; it cannot rewrite code, grant permissions or deploy itself.

03 / ASSURANCE

Tests have a defined scope

Review the developer verification and HTTPS exercise. Developer tests and source/container scanning support this release. They do not establish independent assurance, enterprise accuracy or the absence of vulnerabilities. CodeQL remains skipped because of repository eligibility.

04 / OPERATIONS

Production needs acceptance

The current service is a single-node evaluation on trial hosting. Customer recovery custody, integration acceptance, independent review, sustainable operations and commercial terms remain release gates.

NEXT STEP

Define a small,
measurable evaluation.

Describe your agent framework, one tool, the data labels needed, deployment location and expected allow/deny cases. Share a synthetic example only. Do not send passwords, API keys or customer documents.

A scoped evaluation requires an agreed environment and acceptance criteria. No public self-service account signup, paid plan or production SLA is available yet.

Direct evaluation requests are not open yet. A public business contact and evaluation terms must be supplied by the operator.

Download the buyer brief ↓

Already provisioned? Open the operator console ↗