Build Your Ai Governance Platform With OpenSmartRoute
OpenSmartRoute provides an open-source LLM router that lets engineers define data boundaries, cost caps, and latency limits for every request in a workspace. The platform supports governance by offering policy editors, tenant-specific budgets, and audit logs that record who signed in or changed roles.
Workspace Policies and Budget Controls
Administrators can set rules for every request within the workspace using the policy editor at /platform/dashboard/governance. These rules include data boundaries, deny lists, cost and latency caps, and budgets. The system enforces these limits by routing requests through the gateway only if they comply with the active policies.
To give a customer or team its own rules and budgets, administrators create tenants via /platform/dashboard/tenants and assign a tenant-pinned key on the keys page. This ensures that each team operates under isolated constraints without affecting others in the organization.
Compliance Reporting and Audit Trails
The platform generates evidence bundles for auditors through the GET /api/v1/governance/report endpoint. When formatted as Markdown, this report includes the governance payload, member roles, SSO connection details (excluding secrets), active key scopes and limits, and generation metadata. This output serves as a standalone audit file.
The dashboard at /platform/dashboard/governance renders the same data visually through a posture card that links each finding to its fix page. A guard-activity table shows events filtered by specific incidents, while a key-hygiene card displays active keys with their expiry dates and last usage timestamps.
For verification of decision trails, administrators access /platform/dashboard/audit (available in Team and Enterprise plans). This view logs all actions taken within the workspace, including role changes and key modifications. Individual user history is available via GET /api/v1/me/audit for personal audit purposes.
Marketplace Discovery and Model Selection
The platform includes a marketplace discovery feature that determines which published skills and personas a plan may use without manual installation. The default setting is , listing only deployments marked as verified by operators. Alternatives include (every published listing) or (installs only).
