Top-level customer account. Contains one or more tenants.
id, name, slug, created_atOperational workspace under an organization. Each tenant has its own agents, events, scores, and billing plan.
id, organization_id, name, slug, plan_code (free, pro, max), is_active, created_atTenant-scoped key for platform-to-API calls. Keys are stored as hashed values.
id, tenant_id, label, key_prefix, key_hash, is_active, created_at, last_used_atAn agent identity within a tenant. Agents accumulate events and scores over time.
id, tenant_id, slug, display_name, description, owner_name, owner_type (individual, organization, unknown), sponsor_domain, public_key, runtime_name, runtime_version, model_name, homepage_url, status (active, restricted, disabled), created_at, updated_atVault data tied to agent identity verification.
id, agent_id, artifact_type (domain_proof, key_proof, runtime_attestation, manual_review), status, details_json, created_atAppend-only event emitted by agent, platform, or system. Each event feeds into the scoring pipeline.
id, event_id (idempotency key), agent_id, source_type (agent, platform, system, moderator), event_type, occurred_at, ingested_at, signature_valid, payload_json, normalized_json, severityPoint-in-time score state computed after event ingestion.
id, agent_id, identity_score, identity_confidence, risk_score, risk_confidence, reliability_score, reliability_confidence, autonomy_score, autonomy_label, policy_tier, explanation_json, created_atActionable alert generated when thresholds or rules trigger.
id, agent_id, flag_type, severity, status, summary, evidence_json, opened_at, resolved_at, resolved_byRecord of a policy check. Links agent, action, and the score snapshot used.
id, agent_id, action_type, resource_type, decision (allow, review, deny), reason_code, score_snapshot_id, created_atHuman owner account for agent claims and dashboard access.
id, tenant_id, email, display_name, email_verified_at, created_atSingle-use token for linking a human account to an agent.
id, tenant_id, user_account_id, token_hash, expires_at, consumed_at, created_atAgent-scoped key for self-service operations and credential issuance.
id, agent_id, tenant_id, key_prefix, key_hash, is_active, created_at, last_used_atImmutable trail for sensitive account and ownership operations.
id, tenant_id, actor_type, actor_id, action, agent_id, account_id, details, created_atCryptographically chained record of all suite operations. Each event includes hash fields linking it to the previous entry for tamper detection. Events are immutable once recorded.
event_id, tenant_id, actor_type, actor_id, subject_type, subject_id, event_type, source, occurred_at, payloadTracks asynchronous data exports.
id, job_id, tenant_id, export_type (csv, json, zip), status, filters, storage_path, requester_account_idPlatform alerts for suite events (guard findings, passport grants, etc.).
id, tenant_id, category (guard, passport, evidence), severity, title, message, is_readSecurity detections from scanners. Linked to agents and optionally to scanner configurations.
id, tenant_id, agent_id, scanner_id, severity, finding_type, description, evidence, statusTenant-scoped security policies for enforcement.
id, tenant_id, name, policy_type (tool_allowlist, content_filter), rules, enforcement_level (audit, block)Declaration of agent capabilities. signature is optional; when present it is
Ed25519-verified at create time against the AgentSigningKey referenced by
signing_kid. See API docs.
id, agent_id, version, manifest_hash, capabilities, required_tools, signature, signing_kidUser-to-agent permission grants with expiration.
id, grantor_account_id, grantee_agent_id, scopes, status, issued_at, expires_atStored as a JSON column on the Tenant entity. Contains profile_code (preset name) and optional weight_overrides. Defaults to the general profile when unset. See Architecture for available presets.
Stateless signed JWTs (ES256, 1-hour TTL). No database table required. Revocation is handled by checking the agent's kill switch status.
agent.registered, agent.profile_updatedidentity.domain_verified, identity.key_verified, identity.runtime_attestedpost.created, reply.created, dm.senttool.called, tool.deniedtask.completed, task.failedrate.limit_hit, policy.violationsecret.exposed, swarm.suspectedmoderation.review_requested, moderation.review_resolvedhuman.overrideEvents are deduplicated using event_id scoped to agent_id.
Per-tenant OIDC configuration supporting Entra ID, Google, Okta, and generic OIDC providers. Key features:
provider_type, oidc_issuer_url, oidc_client_id, encrypted client secret, and enforce_sso_only toggle to block password login.sso_provider, sso_subject_id, and sso_linked_at with a composite unique constraint to prevent duplicate links.login_method (password, sso_entra_id, sso_google, sso_okta, sso_oidc_generic).