AI-agent risk intelligence for European organisations
Understand the risks behind your AI agents.
AgentCover helps businesses and AI developers assess, manage and prepare to insure the risks created by autonomous AI agents.
Early-stage product. Workspaces, roles and access control are available today; the risk and insurance modules are in development.
Agent risk overview
12
Registered
9
Assessed
3
High risk
- Procurement Negotiation AgentCritical risk
- Invoice Reconciliation AgentHigh risk
- Support Triage AgentMedium risk
- Sales Outreach AssistantLow risk
Illustrative example with fictional agents. Risk assessment is in development.
- Registry
- Risk
- Insurance
- Monitoring
- Incidents
The emerging problem
AI agents act. Most risk processes assume software only advises.
Autonomous agents now take actions with financial and legal consequences. The tools for understanding that exposure have not kept pace.
Agents hold real permissions
An agent that can send email, move money or change records can cause loss without a person in the loop. Few organisations have a single view of what each agent is allowed to do.
Configurations change quietly
A new tool, a wider data scope or a removed approval step can change an agent's risk overnight. Point-in-time reviews miss it.
Insurance was not written for this
Many policies were drafted before autonomous agents existed. How they respond to a loss caused by an agent is often unclear until someone reads the wording closely.
How AgentCover works
From an inventory of agents to an insurable picture of risk.
Five connected steps. Each builds on the record created by the one before it.
- 01
Register
Record each agent, its purpose, tools, integrations and permissions.
- 02
Assess
Translate configurations into operational and financial exposure, with findings.
- 03
Review insurance
Compare identified risks with existing policies and potentially relevant products.
- 04
Monitor
Track configuration changes and how they move each agent's risk level.
- 05
Respond
Document incidents and collect the evidence an insurer or auditor would ask for.
Main capabilities
One platform, built in stages.
We label what works today and what is planned, so you always know which is which.
Organisation workspaces
Multi-tenant workspaces with roles, invitations and an audit log. Isolation is enforced in the database.
AI Agent Registry
A structured inventory of agents, their configuration, permissions and version history.
Risk Intelligence
Risk scoring, exposure mapping, what-if simulations and recommended controls.
Insurance Intelligence
Policy document analysis, coverage-gap identification and product matching.
Continuous Monitoring
Change detection, alerts and risk history for every registered agent.
Incident and Claims Intelligence
Incident records, evidence collection and reports to support a claim notification.
Who it is for
Built for the teams that deploy agents and the teams that build them.
Businesses deploying AI agents
- See every agent in use and what it is permitted to do.
- Understand operational and financial exposure before it becomes a loss.
- Prepare the information an insurer or broker will ask for.
- Keep a record of changes and incidents.
AI developers and agencies
- Register the agents you build for each client.
- Catch excessive permissions before deployment.
- Produce risk reports that show how an agent was assessed.
- Demonstrate responsible deployment practice to clients.
Why it matters
Responsibility for an agent's actions stays with the organisation.
Delegating a task to an agent does not delegate accountability for the outcome.
- Accountability
- When an agent makes a costly mistake, the organisation that deployed it, and sometimes the firm that built it, answers for it.
- Regulation
- The EU AI Act introduces obligations for providers and deployers of AI systems, phased in over several years. Knowing what your agents do is a precondition for meeting them.
- Insurability
- Insurers price what they can understand. A clear record of agents, permissions, controls and incidents is the starting point for any conversation about cover.
Future insurance partner ecosystem
Designed to work with licensed insurance professionals.
AgentCover is a technology platform. Insurance is arranged and underwritten by licensed brokers, insurers and MGAs, and the platform is being designed so that they can work from the same risk record as their clients.
Where this stands today
- AgentCover has no insurer, broker or MGA partnerships at present.
- It does not sell, issue or underwrite insurance, and cannot confirm whether a policy covers a loss.
- A partner role is reserved in the access model so that authorised partners can be added later.
Security and responsible AI
A risk platform has to be careful with its own.
The foundation is built around tenant isolation and honest product behaviour.
Tenant isolation
Row Level Security on every table. Members see only their own organisation's data.
Server-side authorisation
Roles are checked on the server and again in the database, never taken from the browser.
Audit trail
Invitations, role changes and organisation updates are recorded in an append-only log.
No invented data
Sample data is always labelled as demo data. Unbuilt features are marked as coming soon.
Start with a clear record of your organisation and team.
Create a workspace now and be ready for the agent registry and risk assessments as they are released.