Execution governance for AI
Liberate your AI.
Control and prove every action on your data.
Join the tool call to the query that actually executed. SmartVerify applies your policies and creates one signed record per action, held in your account.
Start record-only. Enable enforcement on your terms.
Tool-server ingress
Person, agent & tool call
Datastore ingress
Executed query & returned data
- Person / agent
- analyst@enterprise / revenue-agent
- Tool call
- get_regional_revenue(region: EU)
- Executed query
- SELECT region, revenue FROM sales…
- Returned data
- 18 rows · region, revenue
- Policy decision
- Allowed · revenue-read policy
Technical alliances
SmartVerify is a member of the Linux Foundation and the Open Secure AI Alliance. SmartVerify is listed as Aligned in the AARM builder registry, a Cloud Security Alliance project.
At the point of execution
A tool call is intent.
An executed query is fact.
Connect the two.
An agent asks a tool to act. The tool reaches your data. SmartVerify intercepts both boundaries to connect the person and agent behind the request to the query that ran and the data that came back.
Paired-interceptor deployment
Inside your environment
01 / TOOL-SERVER INGRESS
Know who asked for what.
Capture the person, verified agent identity, tool call and arguments as the request reaches the tool server.
Agent → interceptor → tool server
02 / DATASTORE INGRESS
Govern what actually happens.
Apply policy on the data path. Capture the executed query and returned data, then join them to the originating tool call.
Tool server → interceptor → datastore
ONE / SIGNED RECORD
Your action. Your evidence.
One signed record per action binds identity, intent, execution and result. The record stays in your account.
How reconciliation worksFits your security stack. Identity, access and agent-platform controls remain essential. SmartVerify complements them with policy enforcement and joined evidence at execution.
Need a narrower starting point? Single-interceptor deployment is also supported for the boundary you need to govern first.
Built by the team that built, operated and scaled security and compliance services at AWS
Four capabilities. Your policies.
Start with visibility.
Move to enforcement.
Begin in record-only mode to understand agent actions on your data. Then enable refusal, reshaping and customer-defined limits where your policies require them.
Record
Join the person, agent, tool call, query as executed, and returned data in one signed entry per action, stored in the customer's account. Retention is configurable, and longer periods with an immutable export are available for estates with a mandated retention period.
Refuse
Refuse a query or tool call that breaks customer policy before it reaches the data or tool server. The agent receives an error its framework already handles.
Reshape
Apply policy to returned data so the agent receives only the rows and fields the request permits. Available on PostgreSQL and MySQL; S3 and DynamoDB reshaping are on the roadmap.
Limit
Cap records per query and tool calls per agent and per person. Customers define the policy limits and counting windows. Behavior-based cumulative limits are on the roadmap.
Available coverage
Record & Refuse: PostgreSQL, including RDS and Aurora; MySQL; S3; DynamoDB; and MCP/JSON-RPC.
Reshape: PostgreSQL and MySQL.
Roadmap · not current capability
Reason and cumulative behavior-analysis limits are on the roadmap. Current limits are defined by customer policy, not learned behavioral baselines.
Built for enterprise ownership
The agent gets access.
You keep the keys.
Govern MCP and JSON-RPC tool calls alongside the datastore operations they trigger. Your policies control the action. Your account holds the evidence.
Explore the platformVerified identities. No shared database secrets.
Short-lived, verified workload identities connect actions to their origin. Agents and tool servers hold no database secrets.
Evidence at the action level.
Each signed action record includes NIST AI RMF and NIST SP 800-53 mappings, giving Security, Risk and Audit teams a common evidentiary trail.
Into the workflows you already use.
SIEM webhooks carry action evidence into your security operations. Investigate with the originating tool call, executed query and result connected.
Let agents work. Keep control.
Put your policies on the path.
Keep the proof in your account.
Start with one workflow in record-only mode. See the action record before you turn on enforcement.