Data retention policy
Conversation history is stored only in memory and automatically deleted after inactivity. Billing metadata (usage counts, costs, balances) is stored indefinitely in our database (until the user requests deletion), but actual message content and AI responses are never persisted. Users can enable Zero Data Retention (ZDR) mode to ensure AI providers also don't store their prompts or responses.
Data archiving and removal policy
In-memory data (conversations, caches) is automatically purged on a rolling basis—conversations after 24 hours, caches after 1 hour. OAuth installation data is automatically deleted when a workspace uninstalls the app. Users can clear their preferences via /ai clear; bulk data export and GDPR deletion requests are handled manually upon request.
Data storage policy
We store only billing metadata (user IDs, token counts, costs, transactions) in Cloudflare D1—no message content, images, or conversation transcripts are persisted. All data is stored in Cloudflare's infrastructure with encryption at rest. User-uploaded images and AI-generated images are passed through to Slack and never stored on our servers.
App/service has sub-processors
yes
Guidelines for sub-processors
App/service uses large language models (LLM)
yes
LLM model(s) used
Users can choose from 300+ models including GPT-5, Claude, Gemini, Llama, and Mistral via OpenRouter; the default is an auto-routing chain that selects the optimal model per request.
LLM retention settings
Zero Data Retention (ZDR) is enabled by default, instructing providers not to store prompts or responses; users can disable ZDR in settings if they need access to models that don't support it.
LLM data tenancy policy
We use OpenRouter as our API gateway with Zero Data Retention (ZDR) enabled by default; individual model providers (OpenAI, Anthropic, Google) maintain their own tenancy policies which users should review.
LLM data residency policy
Data residency varies by model provider—requests may be processed in the US, EU, or other regions depending on which model the user selects; we do not control or guarantee specific data residency.