Agent autonomy.
Human control.

Know your agents. Govern their tools. Keep people in control.

The control layer
for AI agents.

Your agents can read, write, and change systems. Aphura verifies, checks, and records every action.

01 / BEFORE THE ACTION

Establish the identity.

Connect each request to a registered agent, its verified identity, and the organization it belongs to.

02 / AT THE BOUNDARY

Make the decision.

Evaluate policy, tool integrity, sensitive data, and action budgets. Allow, require approval, or block the call.

03 / AFTER THE DECISION

Keep the evidence.

Record decisions, approvals, and outcomes reported by the gateway in a connected audit trail.

See clearly.
Act deliberately.

Identity, permissions, approvals, inspection, budgets, and audit evidence in one control layer.

01

Know every agent.

See registered agents, their owners, verified identities, and the systems they can access.

IDENTITY & INVENTORY
02

Set clear boundaries.

Apply signed policies and scoped permissions to the tool calls your agents make at runtime.

POLICY & PERMISSIONS
03

Keep humans in the loop.

Require an independent person to approve a sensitive operation and its exact arguments.

HUMAN APPROVALS
04

Inspect the exchange.

Check tool definitions, detect sensitive data, and inspect traffic at the MCP gateway.

TOOL GOVERNANCE & DATA CONTROLS
05

Limit the impact.

Set action budgets, investigate exposure, and suspend an agent’s access through Aphura.

BUDGETS & RESPONSE
06

Follow the evidence.

Connect decisions, approvals, and reported outcomes in a tamper-evident audit trail.

AUDIT & INVESTIGATION

Four steps.
Full control.

Connect your agents, choose what they can use, and let Aphura check and record every action.

  1. 01 / CONNECT

    Register your agents.

    Give each agent a verified identity, then route all its tool calls through Aphura.

  2. 02 / GRANT

    Choose its tools.

    Grant included Azure tools or your own MCP servers, with access that expires.

  3. 03 / GOVERN

    Set the rules.

    Apply signed policy and action budgets. Send sensitive calls for human approval.

  4. 04 / AUDIT

    Review the record.

    Trace every decision and outcome in a tamper-evident ledger you can search.

A moment to check.
Before the action.

Let permitted work continue, send sensitive operations for approval, and block unsafe actions.

Your agentVerified identity
AphuraPolicy. Inspection. Control.
Your toolsAuthorized access
THE REQUEST

Read project issues.

A registered agent requests a tool it has permission to use.

Allowed

The call meets policy and stays within the agent’s action budget. Aphura authorizes the handoff and records the reported outcome.

01 / IDENTIFY

Verify who’s asking.

Verify the requesting agent and its organization, then check the identity and permissions behind the action.

02 / EVALUATE

Check the boundaries.

Check signed policy, reviewed tool definitions, sensitive data, and action budgets before allowing a tool call.

03 / AUTHORIZE

Make the decision.

Allow permitted actions, require approval for the exact operation, or block calls that cross a defined boundary.

04 / RECORD

Keep the evidence.

Connect each policy decision and authorized handoff to the gateway’s reported result in the audit trail.

A model included.
Fully governed.

Every plan includes the Aphura Agent Model, powered by OpenAI GPT-6 Luna on Microsoft Foundry in US regions.

  1. 01 / COMPATIBLE

    Keep your code.

    Point any OpenAI-compatible SDK at Aphura and call aphura-agent-1 as is.

  2. 02 / PRIVATE

    Nothing stored.

    Prompts and completions are never stored, and every call stays in US regions.

  3. 03 / GOVERNED

    Checked every time.

    Secrets are blocked, and every model call is recorded in the Aphura audit trail.

Open tools.
Reviewed first.

Open-source MCP servers you run yourself, reviewed before publication and governed on every call.

  1. 01

    Source pinned.

    Each server is tied to an exact, reviewed commit of its public source repository.

  2. 02

    License checked.

    Only MIT or Apache-2.0 servers that you host in your environment are admitted.

  3. 03

    Tools pinned.

    Tool definitions are reviewed and pinned. A changed tool is held for a new review.

  4. 04

    Every call governed.

    The same policy, approvals, budgets, and audit trail apply to every server call.

Every permission.
Every decision.

Govern agent access, review sensitive actions, and investigate every decision in one place.

Control access.

Connect verified agent identities to scoped permissions and signed policies for all approved tool access.

Require oversight.

Require a person to review and approve the exact operation before a sensitive tool request can proceed.

Trace decisions.

Connect requests, approvals, and reported outcomes in an audit trail your team can investigate.

Runs in your Azure.
Stays in your control.

Enforcement runs inside your Azure network, next to your agents. Subscribe through Azure Marketplace.

HYBRID

Enforcement in your VNet.

The gateway and policy checks run in your own Azure virtual network, right next to your agents. Aphura hosts and updates the console, sign-in, approvals, and policy management for you.

SELF-CONTAINED

Everything in your account.

The full platform runs in your own Azure subscription, and the Aphura Agent Model runs on your own Microsoft Foundry endpoint. Only usage reports ever leave your subscription, for billing.

Unlimited agents.
Clear pricing.

Choose your monthly action volume. Every plan includes unlimited agents and the Aphura Agent Model.

STARTING AT

$499/ month

Starter includes 100,000 governed actions per month, shared across every agent in your account.

PAY FOR

Actions

One governed action is one tool call Aphura evaluates, allowed or blocked. No seat charge for adding an agent.

INCLUDED

Model

The Aphura Agent Model is included, with no separate Aphura model charge. Ledger entries are included too.

Keep the control.

Give agents a clear boundary. Give your team a clearer view.

Talk to sales