Tovel brings agent execution, policy checks, human approvals and evidence into one governed environment. See how a workflow runs, where people step in and what gets recorded.
Agents hold identities, act on their own and change state faster than a quarterly review. Tovel gives security, risk, finance and the regulator one place to look. Tovel is not an agent. It does not write code or generate content. It governs the agents that do, and the infrastructure they run on.
The same customer-data export, followed through the three stages every governed action passes: it runs inside an enclave, the controls decide and a person reviews, and the record is written.
The analytics agent runs with the storefront application, its tools and its data inside one governed environment, with a scoped identity and a named owner.
How an enclave is put togetherThe export is checked against policy at the moment it is requested. Bulk sensitive data across a boundary routes to a named approver, who sees the context before anything moves. Denied and expired requests stop here.
The controls in detailActor, policy version, approver, argument fingerprint, outcome and time are written to a signed, hash-chained record. An auditor verifies it against a public key.
Inspect a sample recordAn enclave holds the application, its agents, their tools, the policy that gates them, the people who approve and the evidence that results. Select a view to see the same scenario from each side.
| Agent | Model | Tools | Status |
|---|---|---|---|
refunds-copilotstorefront · T2 | approved model | 6 MCP | Governed |
payments-agentstorefront · T1 | approved model | 9 MCP | Governed |
kyc-revieweronboarding · T1 | approved model | 4 MCP | Governed |
catalog-syncstorefront · T3 | approved model | 3 MCP | Governed |
scraper.unmanagedoff-platform | model unknown | · | Shadow |
Review agents discovered across connected sources, their ownership and governance status. Coverage depends on the sources connected; discovery itself does not block activity.
Risk and control coverage scored from real runtime signals, mapped to the frameworks that apply to you.
Anything above your risk threshold pauses for a named approver, with full context, before it runs.
Every agent-involved action is sealed into a tamper-evident record, ready to show an auditor.
Bedrock and AgentCore supply execution, identity, gateway policy and registry. Tovel adds the operating layer a regulated buyer has to show: named approvers bound to specific actions, versioned organisation rules, a delivery checkpoint and a signed record. You keep the account, the data and the decision.
Security, platform and risk teams each need a different answer first. Each card goes to the part of the site that answers it.
How discovery, policy enforcement and approvals work, and what happens to an action that is denied.
See the controls Platform teamsHow an enclave is put together, what it connects to, and what an evaluation on your own AWS estate needs.
See the enclave Risk and auditA sample evidence record, the three verification checks, and what a record does and does not prove.
Inspect a sample recordEnclaves run in your AWS account on Bedrock and AgentCore, in the region you choose.
Signed, hash-chained receipts exported with their verification material. Retention is set per record type.
Anything above the risk threshold routes to a named person, with full context, before it runs.
On governed paths agents receive least-privilege scopes under a cryptographic identity owned by a named human.
Tovel claims exactly these seven and nothing else. Certifications not held are not claimed. The Trust page lists what Tovel evidences today and what is still in build.
Bring one agent workflow to discuss. We will walk through the relevant controls, show how approvals and evidence work, and identify what an evaluation on your estate would need.
