Crate agenttrustid 

AgentTrust SDK

The AgentTrust SDK provides authentication and authorization for AI agents. It enables secure, auditable agent operations through identity management, capability tokens, pre-flight action checks, and telemetry reporting.

Quick Start

use agenttrustid::{AgentTrustClient, CreateAgentRequest, IssueTokenRequest, AgentTrustGuard};

// Create a client
let client = AgentTrustClient::builder()
    .base_url("http://localhost:8080")
    .api_key("sk_live_xxx")
    .build()
    .unwrap();

// Register an agent
let agent = client.agents().create(&CreateAgentRequest {
    name: "my-assistant".to_string(),
    framework: "langchain".to_string(),
    capabilities: vec!["files:read".to_string()],
    ..Default::default()
}).unwrap();

// Issue an opaque agent token (prefix `at_`)
let token = client.tokens().issue(&IssueTokenRequest {
    agent_id: agent.id.clone(),
    scope: vec!["files:read".to_string()],
    audience: vec!["mcp://filesystem".to_string()],
    ttl: 300,
}).unwrap();

println!("Token: {}", token.token);

Guard Pattern

The AgentTrustGuard provides a high-level wrapper for agents using raw OpenAI/Anthropic SDKs:

use agenttrustid::{AgentTrustClient, AgentTrustGuard};

let client = AgentTrustClient::from_env().unwrap();
let guard = AgentTrustGuard::new(client, "agent-123");

// Before tool calls
guard.check("web_search", "AI news").unwrap();

// After tool calls
guard.report("web_search", true, 1200);

// Telemetry auto-flushed on drop

Re-exports

pub use a2a::A2A;

pub use agentcards::AgentCards;

pub use approvals::ApprovalsAPI;

pub use client::AgentTrustClient;

pub use client::AgentTrustClientBuilder;

pub use delegations::Delegations;

pub use error::AgentTrustError;

pub use error::Result;

pub use federation::Federation;

pub use guard::AgentTrustGuard;

pub use guard::AgentTrustGuardBuilder;

pub use mcp::Mcp;

pub use models::A2ATask;

pub use models::ActionCheckRequest;

pub use models::ActionCheckResult;

pub use models::Agent;

pub use models::AgentCard;

pub use models::AgentCardSignature;

pub use models::AgentCardSkill;

pub use models::AgentExtension;

pub use models::AgentInterface;

pub use models::ApprovalRequestStatus;

pub use models::Capabilities;

pub use models::CreateAgentRequest;

pub use models::CreateDelegationRequest;

pub use models::CreateSIEMDestinationRequest;

pub use models::Delegation;

pub use models::FederationProvider;

pub use models::HealthResponse;

pub use models::InitSessionRequest;

pub use models::IntrospectTokenRequest;

pub use models::IntrospectionResult;

pub use models::IssueFederatedIDTokenRequest;

pub use models::IssueFederatedIDTokenResult;

pub use models::IssueTokenRequest;

pub use models::IssueWIMSETokenRequest;

pub use models::MCPServer;

pub use models::Provider;

pub use models::RegisterFederationProviderRequest;

pub use models::RegisterMCPServerRequest;

pub use models::SIEMDeliveryRecord;

pub use models::SIEMDestination;

pub use models::SecurityScheme;

pub use models::SendTaskRequest;

pub use models::Session;

pub use models::TelemetryEvent;

pub use models::TelemetryReportRequest;

pub use models::Token;

pub use models::UpdateSIEMDestinationRequest;

pub use models::V1Task;

pub use models::V1TaskStatus;

pub use models::VerifyFederatedTokenRequest;

pub use models::VerifyFederatedTokenResult;

pub use models::VerifyWIMSETokenRequest;

pub use models::VerifyWIMSETokenResponse;

pub use models::WIMSETokenResponse;

pub use models::ATI_TRUST_EXTENSION_URI;

pub use sessions::SessionsAPI;

pub use streaming::StreamFilter;

pub use streaming::Streaming;

pub use wimse::Wimse;

Modules

a2a

Agent-to-Agent (A2A) task dispatch via JSON-RPC 2.0.

actions

Pre-flight action authorization checks.

agentcards

Agent Cards API — generate, fetch, and publish A2A-compatible agent cards.

agents

Agent registration and lifecycle management.

approvals

Approval management for elevated actions.

client

The main AgentTrust ID client and builder.

delegations

Capability delegation between agents.

error

Error types for the AgentTrust ID SDK.

federation

Cross-organization OIDC federation.

guard

High-level guard for pre-flight action checks and telemetry.

mcp

MCP server registry and proxy operations.

models

Data models for the AgentTrust ID SDK.

sessions

MCP session management.

streaming

SIEM streaming destination management and event subscription.

telemetry

Telemetry reporting for agent behavior tracking.

tokens

Opaque agent token issuance, introspection, and revocation.

wimse

WIMSE workload identity tokens.