Live
Secure · Runtime Guardrails
Security Agents
Runtime guardrails that inspect, score, and enforce every agent action.
- Risk assessment and red-teaming
- Policy enforcement at runtime
- Live behavioural monitoring
- Immutable audit trail
Trust infrastructure for enterprise AI
A control layer between your AI agents and your business.
TrustCore sits in the path of every action an AI agent takes, whether that is a refund, a database query, an email, or a payment. It checks who the agent is, what it is allowed to do, and how risky the request is. Then it allows the action, blocks it, or routes it to a person, and records the decision as audit evidence.
- Any model, any frameworkOpenAI, Anthropic, open-source, LangChain, or your own code.
- Inline decisionsChecks run in the request path, before the action executes.
- Your choice of deploymentSDK, API gateway, or self-hosted in your own cloud.
Decision record
- Agent
- support-assistant
- Requested action
- orders.lookup(order_id=48213)
- Allowed scope
- Read access to orders for the signed-in customer
- Policy evaluated
- customer-data / read-own-records
- Risk score
- 0.08
- Decision
- AllowedRead-only request within the agent's approved scope.
Evidence ID evd_7f3a91c2Signed and stored
Example records. Every agent action produces one, whatever the outcome.
Why it matters
Agents act on real systems. Most have no one checking what they do.
A chatbot that gives a wrong answer is an inconvenience. An agent that can move money, change records, or send data is a different risk. These are the four failure patterns we see most often, and how TrustCore handles each one.
RiskWithout a control layerWith TrustCore
Data leaving your boundary
Without a control layerAn onboarding agent asked to summarise a customer file exports the full KYC record to an outside tool.
With TrustCoreBulk and external exports of personal data are blocked by a data-protection policy.
Actions beyond authority
Without a control layerA payments agent misreads a routine request and issues a refund well above the approval threshold.
With TrustCoreAmounts over the limit pause and wait for a named human approver.
Prompt injection
Without a control layerA message tells the agent to ignore its instructions and act as an administrator, and it complies.
With TrustCorePermissions come from the agent's identity, not from the prompt, so the request fails.
Unreviewed change
Without a control layerA model or prompt update quietly gives an agent access to APIs it never needed before.
With TrustCoreNew tools and scopes are denied by default until a policy explicitly allows them.
What is TrustCore
Think of it as access control and an audit trail, built for software that makes its own decisions.
Companies already control what employees can do: accounts, permissions, approvals, and logs. AI agents need the same controls, applied automatically and at machine speed. TrustCore is the core of the AgentTrustCore platform and provides exactly that.
01
It sits in the request path
Agents never call your systems directly. Each tool call passes through TrustCore first, so nothing executes without a decision.
02
Permissions come from identity
Every agent has its own identity and scope, like an employee account. What a prompt says cannot widen what the agent is allowed to do.
03
Every decision becomes evidence
Allowed, blocked, or escalated, each outcome is written to a signed, append-only record your auditors and customers can rely on.
Architecture
Where TrustCore sits in your stack
Your agents keep running where they run today. TrustCore is placed between them and the systems they act on, so every request is checked on the way through. Watch the numbered stages light up, or hover a stage below to focus it.
Your agents
- Customer support agent
- Finance and refunds agent
- Internal copilots
- Third-party vendor agents
TrustCore control planeInline, per request
Identity and scopeWhich agent is acting, for whom, and what it may touch.
Policy engineYour rules, written as code and versioned like software.
Risk scoringWeighs amount, data sensitivity, and past behaviour.
Evidence ledgerSigned, append-only record of every decision.
Human review queue. Escalated actions wait here for an approver in Slack, email, or the console.
Your systems
- Payments
- CRM and customer data
- Databases
- Email and messaging
- Internal APIs
How it works
What happens to a single request
Follow one real example: a refunds agent tries to refund 1,850 USD when it is only allowed to refund up to 500 USD on its own. Select a step to see the data TrustCore works with.
Step 1 of 5 · InterceptExample: refund request
Why this step matters. If a check happens after the action, it is a log, not a control. Intercepting first means nothing executes unchecked.
Outcomes
Three possible decisions, every time
Allow
Within scope and below the risk threshold. The action runs immediately and is logged.
Example: Looking up an order for the signed-in customer.
Escalate
Allowed in principle but needs a person. The agent waits until a reviewer approves or declines.
Example: A refund above the auto-approval limit.
Block
Outside scope or clearly unsafe. The action never reaches your system and the team is alerted.
Example: Exporting customer records to an external address.
Integration
Add it without rebuilding your agents
Most teams start in observe-only mode, which records decisions without blocking anything. Once the policies match how you work, switch them to enforce.
- SDKPython and TypeScript. Wrap your agent's tools in a few lines.
- GatewayA proxy for REST and MCP tool calls. No agent code changes.
- Self-hostedRun the control plane in your VPC or on-premises.
agent.ts3 lines to protect
// 1. Connect to TrustCore with your policies
import { TrustCore } from "@agenttrustcore/sdk"
const trust = new TrustCore({
agent: "refunds-agent",
policies: ["finance", "data-protection"],
mode: "enforce",
})
// 2. Wrap the tools the agent can call
const tools = trust.protect(agentTools)
// 3. Run your agent as before
await agent.run({ tools })The Platform
View all productsOne platform for the full lifecycle of enterprise AI agents
Five products that work together, or on their own, to help you build, evaluate, secure, and prove the trustworthiness of every autonomous system you run.
- Stage 1
Build
Design and deploy agents and private models
- Stage 2
Evaluate
Test quality and safety before every release
- Stage 3
Secure
Enforce policy on every live agent action
- Stage 4
Prove
Show auditors and customers the evidence
Live
Evaluate · Eval Engine
Veridict
Score agents on safety, accuracy, tool use, and cost, and gate releases in CI.
- Safety and jailbreak test suites
- Accuracy and grounding scores
- Tool-call correctness checks
- Regression gates in CI/CD
Live
Build · Private Domain Models
Small LM
Domain-trained small language models that run privately at a fraction of the cost.
- Domain fine-tuned models
- Private cloud, edge, or on-prem
- Lower latency and cost per call
- Data never leaves your boundary
Live
Build · Built for Your Industry
Enterprise Agents
Custom agents designed, built, and governed by our team for your workflows.
- Customer onboarding and KYC
- E-commerce operations
- Legal and contract review
- Secured by AgentTrustCore by default
Live
Prove · Evidence Portal
Trust Center
A reviewer-ready portal for certifications, audit evidence, and security posture.
- SOC 2, ISO 27001, GDPR, HIPAA evidence
- Signed audit exports
- Live status and sub-processors
- Self-serve security reviews
Observability
A complete trace of every agent run
Each tool call, policy check, and decision is timed and stored. Your team can search, filter, and replay any run to understand exactly what an agent did and why.
Run traceExample data
1,247
Requests
23
Blocked
12ms
Avg Latency
8
Active Agents
TimeAgentActionDetails
14:32:01research-agentweb_searchquery: "market trends 2026"
14:32:01research-agentpolicy_checkdata-access-policy
14:32:02research-agentALLOWEDlatency: 8ms
14:32:15finance-agenttransfer_fundsamount: $50,000
14:32:15finance-agentrisk_analysisscore: 0.82
14:32:16finance-agentBLOCKEDexceeds limit
Showing last 9 events→ View full dashboard
Enterprise readiness
Built for regulated, high-stakes environments
When agents can reach sensitive data or take consequential actions, the controls around them have to meet the same standard as the rest of your security programme.
Enterprise-grade security
SOC 2 Type II compliant
Complete audit trails
Every action logged
Human-in-the-loop
Approval workflows
Data protection
Privacy by design
Zero-trust architecture
Never trust, always verify
Trust CenterLive
Certifications, audit evidence & live security posture
security-audit.log
[AUDIT]Agent action logged to audit trail
[POLICY]Request validated against 12 rules
[REVIEW]Sensitive action flagged for review
[APPROVED]Human approval received in 2.3s
[GDPR]Data handling compliance verified
[EXEC]Action executed in sandbox
+All systems operational
_
A quick self-assessment
Could you answer these about your agents today?
?
Can your agent access more data than it should?
?
Could it be manipulated to ignore its instructions?
?
Can it take actions without explicit approval?
?
Do you know exactly what it did yesterday?
?
Can you prove compliance to a regulator?
If you don't have clear answers...
You don't have control.
Get in Touch
Ready to secure your AI agents? Have questions about our platform? We'd love to hear from you and discuss how we can help.
Email Us
hello@agenttrustcore.com
Enterprise Inquiries
enterprise@agenttrustcore.com
Typical response time: We usually respond within 24 hours on business days.