Compliance & auditing

Audit trail

Every AI request processed by ModelRiver generates a comprehensive audit record:

• Timestamps: Request received, processing start, provider response, and completion times
• Provider details: Which provider and model processed the request, including fallback attempts
• Token counts: Prompt, completion, and total tokens for cost attribution
• Request and response bodies: The full request and response text, when the project saves it (see Data retention)
• Webhook delivery: Status of all webhook delivery attempts
• Timeline events: Complete request lifecycle including retries and callbacks

Log access

Request logs are available in the console under Request Logs, where you can open each request's timeline and its raw JSON request and response. On Scale and Custom plans, owners and admins can download a project's request logs: open Request Logs, select Download, pick a date range, and choose CSV or JSONL. The download uses the filters on the page. Prompts and responses are left out unless you select Include prompts and responses, and every download is recorded in the organization audit log. A download covers up to 50,000 requests, or with prompts and responses up to 10,000 requests and 200 MB of prompt and response text. For anything else:

• Capture on your side: Store the API responses, webhook payloads, and callback results your application receives
• Contact support: For a specific data request, email [email protected]

Organization audit log

The audit log records who changed what in your organization. Owners and admins on Scale (including Scale+) and Custom plans can browse it in the console (open the organization menu and select Audit log) and download it as CSV.

It records:

• Sign-ins: Successful and failed sign-ins by members of your organization
• Team changes: Invitations, role changes, and member removals
• Projects: Creating, changing, and deleting projects, and turning project protection on or off
• Keys and credentials: API keys and provider credentials being created, changed, or revoked
• Workflows: Publishes and rollbacks
• Webhooks: Webhook configuration changes
• Billing: Plan changes

Each entry shows when it happened, who did it (a member, an API key, or ModelRiver itself), what changed, and whether it succeeded.

Every organization's activity is recorded on every plan, so when you upgrade to Scale you can see the past 12 months right away. Entries are kept for 12 months (24 months on Custom). The IP address and browser on each entry are cleared after 90 days. Entries can't be edited or deleted, and they never contain API keys, passwords, prompts, or AI responses.

Access controls

• Organization roles: Owner, Admin, Editor, and Viewer roles control what each member can see and change (see Organization roles and project protection)
• Single sign-on: On Scale and Custom plans, members sign in through your OpenID Connect identity provider, and the owner can require it (see Single sign-on)
• Per-key isolation: Separate API keys per environment prevent cross-contamination
• Session security: Dashboard sessions use secure, HttpOnly cookies with configurable timeouts

Compliance checklist

Least privilege principle

Maintain security by following least privilege:

• Issue dedicated API keys for each service or team
• Use short-lived keys (1-day or 7-day expiration) for development environments
• Revoke keys immediately when team members leave or integrations are decommissioned
• Review active keys periodically in Settings → API Keys

Next steps

• Input guardrails: Gateway content policy enforcement
• API keys: Manage authentication credentials
• Provider credentials: Secure AI provider tokens
• Observability: Monitor and audit request logs