Delivering trusted AI agent and MCP server identity for secure, accountable, autonomous systems

New SaaS application demonstrates how to link to Cisco Duo, Okta or ORY identity providers to establish trust for MCP servers, A2A, and OASF agents.

As AI agents become integral to enterprise workflows, securing their identities and actions has emerged as a critical trust challenge. Unlike humans or static applications, autonomous agents operate at machine speed, shift roles instantly, and may exist only for the lifespan of a single task.

Traditional identity systems weren’t built for this reality. They falter at enforcing fine-grained permissions, ensuring clear attribution, and safeguarding sensitive credentials — leaving dangerous gaps in control, accountability, and safety. 

The AGNTCY Agent Identity framework is purpose-built to meet this challenge head-on. It is specifically designed to keep pace with ephemeral agents who are autonomous, operate across organizations and adapt quickly. 

The framework ensures that every AI agent can be authenticated, tracked, and trusted before taking any action. Built as part of the AGNTCY open source project that is focused on tackling key challenges around agent identity as well as agent discovery, messaging, observability and evaluation, the Agent Identity framework is now available as a free SaaS application from Outshift by Cisco. 

The Outshift Agent Identity Service powered by AGNTCY helps users learn how to establish a secure and verifiable identity for AI agents, multi-agent services, and Anthropic’s Model Context Protocol (MCP) servers. The service offers organizations the opportunity to define and test an agent identity strategy without having to first invest in building and deploying their own. 

Read on to learn more about the Outshift Agent Identity Service and what we see coming next in the realm of agentic identity.

The Outshift Agent Identity Service: Easy-to-use identity services for MCP servers, A2A, and OASF agents 

Outshift Agent Identity Service powered by AGNTCY is a free SaaS application that demonstrates how the AGNTCY Agent Identity framework can manage verifiable identities and access control for AI agents, multi-agent services, and MCP servers. 

The service allows users to register and verify identities, issue trusted badges, and define fine-grained access control policies — all from one place. Using an intuitive dashboard or API, developers can issue trusted agent badges, enforce scoped permissions, and manage agent-tool interactions. 

After verifying the identities of AI agents and/or MCP servers, organizations can leverage these agentic services to address a range of critical use cases, such as:

1. Ensuring AI agents in a retail chain can only place orders through verified MCP servers connected to authorized suppliers.
2. Preventing AI agents in doctor’s offices from sharing patient records with unverified or unauthorized external systems.
3. Enabling AI agents to handle more customer service interactions by securely accessing back-office systems and trusted enterprise knowledge bases through MCP servers.

By combining identity assurance with policy-driven access, organizations are able to enjoy stronger security, compliance alignment, and streamlined agent operations.

Key features

1. On-demand badge generation – Instantly create and preview verifiable badges for agentic services (AI agents, MCP servers) that follow a variety of specifications, including Google’s Agent2Agent (A2A), MCP and Open Agentic Schema Framework (OASF).
2. Fine-grained control – Create and enforce fine-grained access control policies for agentic services.
3. Human-in-the-loop approvals – Add an extra layer of protection to sensitive actions by creating policies requiring real-time human authorization.
4. Flexible issuers – Tap into your trusted Cisco Duo, Ory or Okta Identity Provider for new identities, or issue verifiable, decentralized identities directly through AGNTCY’s IdP.
5. Device onboarding – Register and manage personal devices to enable secure authentication and receive identity approval notifications for human-in-the-loop approvals.
6. Graphical user interface – An easy, intuitive dashboard allows users to manage agent identity through the full lifecycle — registration, badge creation, and identity verification.
7. Python and gRPC APIs/SDKs – Integrate identity and policy management into your workflows with endpoints for Agent Directory, MCP servers, A2A agents, and OASF systems.

The Agent Identity Service standardizes identity for MCP, A2A, and OASF ecosystems using verifiable, cryptographic badges — delivering trust, interoperability, and policy control across your agentic environment. 

Learn more about how to get started using the service by reading the documentation or attending our upcoming webinar, The Key to Autonomous AI Agents and MCP Servers You Can Trust.

Example use case: Secure currency exchange that uses Cisco Duo, Okta or ORY identity provider

We built a multi-agent currency exchange application to show how the Outshift Agent Identity Service delivers secure AI agent identity, fine-grained access control, and trusted communication between agents and servers. 

In this example application, a large retail bank offers customers a financial assistant chat that can provide information on currency exchange rates and assist with instant currency exchanges. Behind the scenes, this service relies on multiple AI agents and an MCP server — all registered, verified, and governed by the Agent Identity Service to ensure only authorized actions occur and to secure every interaction within the workflow (See: currency exchange samples).

Currency exchange software components: A2A, MCP and OASF agentic services

Five steps to onboarding AI agents, multi-agent services, and MCP servers

1. Sign up and create an organization – Set up your organization account in the service.
2. Connect identity provider – Link Cisco Duo, ORY, Okta, or use the built-in demo AGNCTY IdP.
3. Onboard devices – Register and manage devices for secure authentication, human-in-the-loop approvals, and push notifications.
4. Register and badge – Add your agents, multi-agent services and MCP servers, then issue them verifiable badges.
5. Verify, configure, and embed – Validate badges, retrieve API keys/tokens, embed them into agents, servers, and enable human-in-the-loop approval flows where required/desired.
6. Set policies and go live – Define tools and permissions that can be accessed by agentic services, then run with secure, policy-driven access and real-time human authorization for sensitive actions.

Securing the currency exchange workflow

Here’s how the Outshift Agent Identity Service secures the currency exchange workflow:

1. User request – The customer types “Convert 100 USD to EUR” in the financial assistant chat.
2. Authenticate and policy check – The financial assistant agent (OASF) authenticates with the IdP and confirms it has permission to start the workflow with the currency exchange (A2A Agent) and/or the currency exchange (MCP Server).
3. Agent authorization– The financial assistant agent uses the API key to call the currency exchange and/or MCP server. Once the Outshift Identity Service validates the identity and verifies that the financial assistant agent has authorized access, the workflow can continue.
4. Human approval via mobile device – When a sensitive request is made, the service enforces policy by sending a live approval notification to an authorized approver’s mobile device. The process continues only after explicit confirmation.
5. Identity and device trust – Validate identities, enforce policies, and confirm trusted devices.

Advancing AI agent identity towards Zero Trust

The launch of Outshift Agent Identity Service powered by AGNTCY marks a pivotal step toward securing autonomous AI agents at scale. 

This service offers easy-to-use interfaces for establishing verifiable identities, defining scoped permissions, and enabling interoperability across MCP, A2A, and OASF ecosystems. But this is just the beginning. We envision that, over time, identity will evolve into a more dynamic trust signal — continuously verified and contextualized — to define, enforce, and validate trust for every agent action. 

This transformation will move agent security from reactive defense to proactive governance, empowering enterprises to innovate with confidence while maintaining operational integrity.

Learn more about how we’re building this trust-first agentic future — register for our upcoming webinar to see how the Outshift Agent Identity Service and Zero Trust principles can secure autonomous systems from day one.

Watch this workflow in action: See the currency exchange demo on the Outshift YouTube channel.