TypeScript Types

The SDK re-exports commonly used types from @agenthifive/contracts so you can import everything from a single package. These types are used throughout the SDK's method signatures and return values.

Importing Types

import type {
  Connection,
  ConnectionStatus,
  OAuthProvider,
  Agent,
  Policy,
  ExecutionModel,
  DefaultMode,
  StepUpApproval,
  AllowlistEntry,
  RateLimit,
  TimeWindow,
  AuditEvent,
  AuditDecision,
  ExecuteRequest,
  ExecuteRequestModelA,
  ExecuteRequestModelB,
  ExecuteResponse,
  ExecuteResponseModelA,
  ExecuteResponseModelB,
  ExecuteResponseApproval,
} from "@agenthifive/sdk";

Core Entity Types

Type	Description
Connection	A connected OAuth provider account with token storage metadata.
ConnectionStatus	Union of connection lifecycle states (e.g., "healthy", "needs_reauth", "revoked").
OAuthProvider	Supported OAuth provider identifiers: "google", "microsoft", "telegram", "github", "slack", "anthropic", "openai", "gemini", "openrouter", "notion", "trello", "jira".
Agent	A registered AI agent or application within a workspace.
AgentStatus	Agent lifecycle state: "created", "active", or "disabled".
Policy	An access policy binding an agent to a connection with rules.

Policy Rule Types

Type	Description
ExecutionModel	Execution model identifier: "A" (token vending) or "B" (brokered proxy).
DefaultMode	Default policy behavior: "read_only", "read_write", or "custom".
StepUpApproval	Step-up approval configuration: "always", "risk_based", or "never".
AllowlistEntry	URL pattern (baseUrl, pathPatterns) and HTTP methods permitted by a policy.
RateLimit	Rate limiting: maxRequestsPerHour, optional maxPayloadSizeBytes and maxResponseSizeBytes.
TimeWindow	Time-based access restrictions (e.g., business hours only).

RequestRuleAction

Union of actions a request rule can take: "allow", "deny", "require_approval", "redact".

RequestRule

A request rule evaluated by the policy engine (first-match wins).

Property	Type	Description
label	string?	Optional human-readable label for the rule.
match.methods	HttpMethod[]?	HTTP methods to match (e.g., ["GET", "POST"]).
match.urlPattern	string?	URL pattern to match against the request URL.
match.queryPattern	string?	Query string pattern to match.
match.body	BodyCondition[]?	JSON body conditions (path + operator + value).
match.pii	PiiMatchConfig?	PII detection config -- scans request body fields for PII without mutating the payload.
action	RequestRuleAction	Action to take when the rule matches.
redactConfig	RedactConfig?	PII redaction config -- only used when action is "redact". Specifies which PII types to detect and which body fields to walk.

RedactPattern

Specifies a PII pattern type for redaction. Has a type field (see below), an optional pattern (for "custom" type), and an optional replacement string.

Groups (expand to multiple recognizers): "all_pii", "financial", "identity", "contact"

Generic patterns: "email", "phone", "credit_card", "iban", "ip_address", "url", "crypto_wallet", "date_of_birth", "mac_address", "secret_code"

US-specific: "us_ssn", "us_itin", "us_passport", "us_driver_license", "us_bank_routing", "us_npi"

UK-specific: "uk_nhs", "uk_nino"

Italy-specific: "it_fiscal_code", "it_vat", "it_passport", "it_identity_card", "it_driver_license"

India-specific: "in_aadhaar", "in_pan"

Spain-specific: "es_nif", "es_nie"

Australia-specific: "au_tfn", "au_abn"

Other countries: "pl_pesel", "fi_pic", "th_tnin", "kr_rrn", "sg_fin"

Legacy alias: "ssn"

Custom: "custom" -- requires a pattern (regex) field.

Execution Types

Type	Description
ExecuteRequest	Union of Model A and Model B execution requests.
ExecuteRequestModelA	Request shape for token vending (Model A).
ExecuteRequestModelB	Request shape for brokered proxy (Model B) -- includes method, url, query, headers, body.
ExecuteResponse	Discriminated union of all execution response types.
ExecuteResponseModelA	Response from Model A: contains accessToken, tokenType, expiresIn.
ExecuteResponseModelB	Response from Model B: contains status, headers, body.
ExecuteResponseApproval	Response when step-up approval is required: contains approvalRequestId.

Audit Types

Type	Description
AuditEvent	A single audit log entry with timestamp, action, decision, and metadata.
AuditDecision	The outcome of a policy evaluation (e.g., "allowed", "denied", "error").

SDK-Specific Types

These types are defined in the SDK itself (not re-exported from contracts):

import type {
  AgentHiFiveClientConfig,
  ConnectionSummary,
  ConnectStartResult,
  ExecuteModelAResult,
  ExecuteModelBResult,
  ExecuteApprovalResult,
  ExecuteResult,
  ExecuteModelBOptions,
  ApprovalRequest,
  AgentSummary,
  CreateAgentOptions,
  CreateAgentResult,
  BootstrapResult,
  TokenResult,
  PolicySummary,
  CreatePolicyOptions,
  AuditListOptions,
  AuditListResult,
} from "@agenthifive/sdk";

Type	Description
AgentHiFiveClientConfig	Constructor options for AgentHiFiveClient (baseUrl, privateKey+agentId or bearerToken).
ConnectionSummary	Simplified connection object returned by listConnections().
ConnectStartResult	Result of connect() containing the authorizationUrl.
ExecuteResult	Discriminated union of ExecuteModelAResult, ExecuteModelBResult, and ExecuteApprovalResult.
ExecuteModelBOptions	Options for Model B execution: connectionId, method, url, query, headers, body.
ApprovalRequest	Enriched approval request with agent and connection metadata.
AgentSummary	Agent with status (AgentStatus) and enrolledAt (nullable).
CreateAgentResult	Created agent with bootstrapSecret (shown once).
BootstrapResult	Result of AgentHiFiveClient.bootstrap(): agentId, name, status, workspaceId.
TokenResult	Token exchange result: access_token, token_type, expires_in.
PolicySummary	Policy object with resolved allowlists, rate limits, and time windows.
AuditListOptions	Filter and pagination options for listAuditEvents().
AuditListResult	Paginated audit event response with events array and nextCursor.
ServiceInfo	Service descriptor returned by listServices(): id, name, provider, icon, and actions array.
ServiceActionInfo	Action within a service: id, label, description, requiresApproval.
MyCapabilities	Agent capability status returned by getMyCapabilities(): activeConnections, pendingRequests, availableActions.
ActiveConnection	An active connection with connectionId, service, label, and actionTemplateId (nullable).
PendingCapabilityRequest	A pending capability request: id, actionTemplateId, reason, requestedAt.
AvailableAction	A discoverable action: id, serviceId, label, description, requiresApproval.
CapabilityRequest	Result of requestCapability(): id, actionTemplateId, reason, createdAt.