acAgentCitiby TYRTalk to Us
Concept illustration of a transparent AgentCiti, connecting people, agents and institutions

The Next Leap:
From Digitization
to Agentization.

Build Your AgentCiti.
Shape the Agent Economy.

AgentCiti is designed to connect agents and services—turning everyday needs into coordinated action, with people shaping the rules.

Talk to Us See It in Action
01 / A connected world of agentsConcept illustration

People operate software People delegate work.

Agent security, evaluation and governance by TYR.

02 / AgentCiti in everyday life

One request.
A whole journey.

A personal agent connects the services around you.
Less time coordinating. More room for life.

Connected through AgentCitiIllustrative journey

“Plan my trip.”

Your personal agent

Your preferences guide the journey.

  • Flights & transportFind a way there
  • Hotels & staysFind a place to stay
  • Entry supportUnderstand entry requirements
  • PaymentsPlan how to pay
  • Local careArrange support, if needed
What this means for you

A trip that starts
with you.

Share where you want to go, when, and what matters to you. Your personal agent brings that context to the services you need.

Your destination
Where you want to be
Your preferences
How you like to travel
Your budget
The limits you choose

One request. A shared starting point.

For peopleMore of the journey handled together.

For service providersNew connections to customers and partners.

Explore the agent workspace

03 / See & align

A Lens into
Your Agent World.

Seeing Is Believing.

See clearly. Align actions. Set the pace.

AgentCiti is designed to make connected assets, agent actions and their effects visible and verifiable. Bring people and agents into shared context, at a pace people can understand.

Inside your AgentCiti
Interactive illustration · Sample data
Shared task MAINT-042 / context v1
DEVICE MAINTENANCEReady to begin
No live systems connected
01Inspect
02Plan
03Review
04Approve
05Verify
01 / SEE

Real assets.
Verifiable actions.

Connect what appears in the City to identifiable devices, systems and evidence you can inspect.

02 / UNDERSTAND

Bring what matters
into focus.

Make progress, failures and detected conflicts visible, so people can judge what needs attention.

03 / ALIGN

Shared context.
Across boundaries.

Align people, agents and organisations around shared tasks, with the information each role may access.

04 / DIRECT

Work at
your pace.

Adjust supported workflows, with checkpoints and pauses when people need to decide or intervene.

Visibility and control follow your permissions and the systems connected to your City.

04 / Create value

Unlock Value You
Haven’t Yet Imagined.

Build on what you have.
Expand what you can do.

01

Deliver more with
existing resources.

Put your systems and expertise to work across more tasks, with fewer handoffs.

Greater capacity
02

Reach people through
their agents.

Help customers and citizens access your services through the agents they authorise.

Wider reach
03

Create new services
with partners.

Combine complementary capabilities to develop joint services and explore new revenue opportunities.

New possibilities for growth

05 / The TYR foundation

Trust at the
point of action.

Agent security, evaluation and governance, designed for work that crosses systems and institutions.

01PROTECT

Agent
Security

Verify whom an agent represents. Define what it may access and do. Apply security and permission checks before protected actions execute.

Identity & authorityPermission checks
02EVALUATE

Multi-Agent
System Evaluation

Evaluate individual agents and multi-agent systems through security testing and benchmarking. Assess task performance, coordination and resilience, including adversarial inputs, conflicting objectives and failure recovery.

Security testingBenchmarkingSystem behaviour

Test before deployment. Refine evaluations with authorised operating evidence as systems evolve.

03GOVERN

Governance
in Action

Let people set the boundaries for autonomous work, hold consequential steps for review and intervene within their authority, with records they can verify.

Human approvalIntervention & evidence

Information can inform an agent.
Authority comes from people and institutions.

06 / Digital capital & sovereignty

Your digital capital.
Your rules.
Your future.

Your agents, data, knowledge, skills and workflows are the foundations of lasting capability. AgentCiti is designed to help you build on these assets and govern how they are accessed, used and shared.

Each institution governs its own environment. Connections between AgentCiti environments follow mutually agreed rules, with each participant retaining authority over its own assets and decisions.

We build the foundation.
You govern your AgentCiti.
Your digital territoryBuilt by you.
Governed by you.
01Agents02Data03Knowledge04Skills05Workflows06Networks

Control how your assets are accessed,
used and shared.

07 / Our vision

The Agent Economy
should work for more people.

Our ambition is a connected world of independently governed AgentCiti environments—where local businesses can reach new customers, public institutions can extend their services, and communities can bring their knowledge and capabilities to wider opportunities.

Concept illustration of three independently governed AgentCiti environments connected through shared gateways
Independent institutions. Agreed connections.Concept illustration

The AgentCiti Manifesto

Let intelligence serve people.
Let action be governed.
Let no one be left behind.

Cooperation should create value. Competition should reward useful, accountable performance. People should have a say in how their agents act and how their knowledge contributes to that value.

They should have the context and time to make decisions that matter. Participation should be accessible across ages and abilities. Essential services should remain available to people who do not use agents.

08 / Build with us

Start with one workflow.
Build what comes next.

Begin with an Enterprise Lab or City Lab: define a real task, connect the right participants, test the boundaries and measure the outcome. Build reusable capabilities and a local team that can operate and extend them.

Talk to Us
contact@tyr.ai
OUR FOUNDING VISION

AgentCiti Manifesto

Let Intelligence Serve People. Let Action Be Governed. Let No One Be Left Behind.

We are beginning to hand parts of our lives to AI agents: booking appointments, paying bills, filing forms and dealing with other people’s agents. A chatbot gives you an answer. An agent does something, and what it does lands on real accounts and real people.

Large language models (LLMs) supply the intelligence. Agents turn it into action. Ask an LLM to read an invoice and it will tell you what the invoice says. Give it access to accounts payable and it can change the supplier's bank details and release the payment. The intelligence is the same. What has changed is what it has been allowed to touch. Once an agent can act on someone's money or records, it matters who the agent is, what it may do, and who answers for it when something goes wrong.

The LLMs behind an agent may remain in the hands of a few large companies. But the agent itself need not. What it remembers, what it may access and whether it can move with you between platforms should be your decisions. Otherwise, a personal agent is only another rented service.

AgentCiti is not a city governed by AI. It is a city that governs AI agents. Intelligence may come from many suppliers. The right to decide what an agent may do in your name stays with you, or with the organisation it acts for.

What Is AgentCiti?

A resident’s agent may ask a hospital to move an appointment, a bank to settle a bill and a municipal office to update a record. Each institution keeps its own systems and data. Their agents should be able to work together without forcing everyone onto one platform. This is why the city, rather than the platform, is the right place to set the common rules.

AgentCiti is a city-scale operating and governance framework for multiple AI agents in the real world. Each person, business and public body keeps its own agents, its own data and its own internal rules. One rule applies across the city: whenever an agent enters another party’s system, the interaction must pass through a secure gateway at that boundary. Each institution runs its own gateway at its system boundary, using a common standard that different providers can implement.

Before an action goes ahead, the gateway verifies the agent and whom it represents, grants only the permissions needed for the task, records the information and tools used for the action, and applies the rules on both sides of the boundary.

The city maintains a visible map of agents acting in public and institutional systems. Agents appear where they act, linked to the people or organisations they represent. Participating institutions set access to their records under rules agreed with residents. Within those limits, residents, administrators and regulators can see who acted, where and under whose authority. Private conversations, memories and everyday choices stay off the map. The gateway enforces the rules; the map shows who holds what authority.

What AgentCiti Must Do

Benefits reach everyone. AgentCiti should be judged by how many people gain the ability to get things done. An older resident might use her agent to deal with the clinic and the electricity company, within a spending limit she set herself. A shop owner might let her agent order stock below a set price and answer customers, but never access payroll. Neither of them should have to learn to program, write prompts or give up their data to get the service. Someone who cannot set up an agent should still be able to use one provided by a relative, a school or a public office, within limits the user controls.

Agents move with their users. If a family has spent three years teaching an agent how their household works, that knowledge should go with them when they change platforms, just as a phone number can move between networks. An agent's identity and its memory belong to its user. And an organisation should be able to check who an incoming agent is and what it may do without everyone joining the same platform.

Only the owner sets an agent's authority. Before an agent acts, its authority must be clear and limited: what it may do, what data and tools it may use, how long access lasts and, where money is involved, whom it may pay and how much. Information is not an instruction. External content may inform an agent, but it cannot expand or redirect the authority granted by people and institutions. The gateway checks consequential actions against the rules on both sides of the boundary, escalating or blocking them where necessary, and leaves a record that can be checked.

Start with People. Begin in Real Cities.

AgentCiti is a city for people. An agent may act within delegated authority, but responsibility remains with people and institutions. When an agent's action affects someone's rights, that person must be able to find out what happened, object, and have a human being review it. And nobody should lose access to a public service because they do not have an agent, cannot use one or would rather deal with a person.

AgentCiti does not have to begin with a whole city. It can begin with one hospital, one school or one district office, a handful of agents and one real task crossing organisational boundaries. The test should be plain: Did it work better? Was anyone’s data exposed or an existing service lost? Did people who were previously excluded gain access? Only rules that pass this kind of test in a real place should become the rules of a city.

No company or government can build AgentCiti alone. Rules for admitting agents into a city’s systems must be written with the institutions and residents who will live with their consequences. The people who own the largest models will have a say in how AI is used. So should the people who will feel its consequences.