SmartRetro’s AI features — summaries, clustering, the deeper pattern work — can run under your own LLM provider key instead of ours. Model calls then happen on your account, which many teams prefer for control and auditability.
Adding a key
An admin goes to Settings → Integrations and opens the AI provider card. Choose a provider — OpenAI or Anthropic — and paste your API key. You can optionally name a specific model; leave it blank to use the provider’s default. SmartRetro validates the key on save and, once validated, shows the provider, the model, the last four characters of the key, and when it was validated. The full key is stored encrypted and never shown again.
What runs on your key
With a validated key in place, SmartRetro’s model calls for your organization run through your provider account. Because the calls are made under your own account, you can audit them on the provider’s side directly. For the full picture of what data is sent and what’s stored, see Trust & security.
What happens without a usable key
AI is a layer on top of the rule-based core, not the whole product. If there’s no usable key — none added, or one that’s failing — on a plan where AI would otherwise run, the AI features pause. The rest of the loop keeps working: rule-based signals still flow, retros still run, action items still track. When AI is paused, the app says so plainly rather than failing quietly.
Removing a key
Remove the key from the same card. AI features then fall back to your plan’s defaults. Removal takes effect immediately.
Next
See Signals and The brief for the surfaces AI contributes to, and Organization settings and billing for how plans decide what’s available.