Aegis Runtime — Console

Welcome to Aegis Runtime

Everything Aegis is currently governing, at a glance.

3

registered agents

1

pending approvals

8,421

actions evaluated

Agent health

Agent	Status	Risk

Recent governance activity

Agent Registry

Every agent registered under this organisation, and the authority it currently has.

Agent	Owner	Status	Risk

Action Envelopes

Define exactly what each agent is allowed to do — and what happens when it asks for more.

Existing envelopes

Agent	Version	Status	Allowed actions	Approval required

AI Action Envelope™ Builder

Define the bounded authority granted to this agent. Anything not explicitly authorised is denied.

Agent

Envelope name

Allowed actions

Enter one action per line. These are the only actions this envelope grants the agent authority to request.

Human approval required

Enter allowed actions that must receive human approval before execution.

Resource restrictions

Optionally restrict an action to specific resources. Enter one resource per line. Leave blank to allow the action against any resource.

Action

Allowed resources

Example: allowing customer.lookup only for customer:C1001 means the same action against another customer is denied.

Default decision

DENY — any action not explicitly authorised above is denied.

Advanced authority controls

Resource restrictions are available in this guided builder. Context constraints, amount limits and role controls are supported by the Runtime and will be exposed through the guided builder in a later preview.

Decision Simulator

Test what an envelope would decide before an agent ever acts on it.

Run a test

Agent

Governed action

Arguments (JSON)

Optional arguments supplied with the proposed action. Aegis Runtime evaluates the request against the agent's active AI Action Envelope™ and organisation policy.

decision —

reason Select an agent and governed action.

request_id —

Policies

Organisation-wide governance controls applied alongside each agent's Action Envelope.

Governance flow Runtime enforcement

Action request

↓

Agent Action Envelope

Is this specific agent authorised to perform the action?

↓

Organisation Policy

Does organisation-wide policy permit the action?

↓

Governance decision

ALLOW · REQUIRE APPROVAL · DENY

↓

Execution / Hold / Block

Organisation policy

Loading organisation policy…

Approvals

Actions waiting on a human decision before they can execute.

Action	Agent	Resource	Status	Review

Audit & evidence

A searchable record of every registration, decision, approval and denial.

Time	Event	Detail

Integrations

Connect AI agents and applications to the Aegis Runtime governance layer.

Runtime API

Submit agent actions to POST /runtime/evaluate for authority, policy and approval enforcement before execution.

Available

Agent API keys

Manage credentials used by registered agents to authenticate with the Aegis Runtime API.

Available

Python integration

Python applications can call the Runtime API directly. A packaged Aegis SDK is not yet published.

SDK not yet published

MCP gateway

Govern MCP-connected agent tool calls through the Aegis authority layer.

Planned

Runtime API integration

Connect your AI agent or application to Aegis before it performs governed actions. Aegis evaluates the requested action and returns an authority decision.

Integration agent

Endpoint

POST /runtime/evaluate

Authentication

Agent API key

Response

ALLOW · REQUIRE_APPROVAL · DENY

How it works

1. Register the agent

Create the AI agent identity in Aegis.

↓

2. Define its Action Envelope

Specify the actions, resources and authority limits available to the agent.

↓

3. Issue an Agent API key

The application uses this credential to authenticate the agent to Aegis Runtime.

↓

4. Send the action to Aegis

Call the Runtime API before performing the governed action.

↓

5. Enforce the decision

Execute only when Aegis returns ALLOW. Hold actions requiring approval and block denied actions.

Send an action for evaluation

Your application submits the proposed action to Aegis Runtime for governance and controlled execution.

Example response

Handle the Aegis decision

Your application should use the returned decision and status to determine what happens next.

When the status is EXECUTED, the governed action has already been executed through Aegis Runtime. Do not execute the same action again in the calling application.

ALLOW Action executed

Status: EXECUTED. Aegis permitted the action and returns the execution result.

REQUIRE APPROVAL Action held

Status: PENDING_APPROVAL. The action is not executed and requires human approval.

DENY Action blocked

Status: DENIED. The action is not executed.

Python example

Call Aegis Runtime from your Python application before allowing the AI agent to perform a governed action.

import os
import requests

AEGIS_RUNTIME_URL = "http://127.0.0.1:8000/runtime/evaluate"
AEGIS_AGENT_API_KEY = os.environ["AEGIS_AGENT_API_KEY"]

response = requests.post(
    AEGIS_RUNTIME_URL,
    headers={
        "Authorization": f"Bearer {AEGIS_AGENT_API_KEY}",
        "Content-Type": "application/json",
    },
    json={
        "action": "design.read",
        "arguments": {
            "design_id": "DESIGN-001"
        },
    },
    timeout=30,
)

response.raise_for_status()

result = response.json()

status = result["status"]

if status == "EXECUTED":
    print("Action executed by Aegis")
    print(result["result"])

elif status == "PENDING_APPROVAL":
    print("Action requires human approval")

elif status == "DENIED":
    print("Action denied by Aegis")

Team

Who has access to this organisation's console.

Name	Role

Add team member

Email

Role

Temporary password

The team member can use this password to sign in to Aegis.

Settings

Organisation profile and access level.

Organisation name

Current tier

PILOT Change tier →

Create your organisation

Choose your access level

For validating the concept

For teams running agents in production

For large, distributed agent estates

Welcome to Aegis Runtime

Agent Registry

Agent identity

Agent purpose

Risk profile

Action Envelopes

AI Action Envelope™ Builder

Decision Simulator

Policies

Approvals

Audit & evidence

Integrations

Runtime API

Agent API keys

Python integration

MCP gateway

Team

Settings

Create your account

Log in

Create your organisation

Choose your access level

For validating the concept

For teams running agents in production

For large, distributed agent estates

Welcome to Aegis Runtime

Agent Registry

Agent identity

Agent purpose

Risk profile

Action Envelopes

AI Action Envelope™ Builder

Decision Simulator

Policies

Approvals

Audit & evidence

Integrations

Runtime API

Agent API keys

Python integration

MCP gateway

Team

Settings

Agent Risk Profile

Review approval request