FAQ

Common questions about rendfly — pricing, privacy, providers, retention.

If your question isn’t here, email hi@rendfly.com.

Where are conversations stored?

Conversations are stored in Postgres and ClickHouse, both encrypted at rest (AES-256-GCM on sensitive columns including conversation content, system messages, and provider keys). We host in EU and US regions; the region is pinned per workspace at creation and doesn’t change without your explicit request. You can ask for the specific region during sign-up.

How is PII handled?

PII redaction runs before storage. Names, email addresses, phone numbers, and physical addresses are tokenized on ingestion — the raw values are not written to the conversation store. The original, pre-redaction text is held only for a short purge window (configurable per workspace, default 24 hours) and then hard-deleted. You can shorten that window to zero in your workspace settings if your data policies require it.

Do you train on my data?

No. Customer conversations and system messages are never used to train or fine-tune any model. They flow only into the judge calls you’ve configured for your own projects. rendfly’s judge uses third-party model providers; those calls are made under rendfly’s account by default (or your own on Enterprise with BYOK), but neither rendfly nor the underlying providers use your data for training under our agreements.

What’s your compliance status?

LGPD and GDPR controls are in place from day one — PII redaction, data minimization, short retention windows, and customer-controlled deletion. SOC 2 Type I is targeted for Q4 2026, SOC 2 Type II for Q1 2027. ISO 27001 will follow. If you need specific controls documented before those certifications land, email hi@rendfly.com and we’ll share the current control evidence directly.

Which LLM providers are supported?

Proxy mode is provider-universal: any provider that accepts OpenAI-compatible endpoints works — OpenAI, Anthropic, Google, Groq, Mistral, vLLM, Ollama, and others. You change one line in your agent code (base_url) and rendfly intercepts the traffic. API mode covers OpenAI Stored Completions (read-only admin key, zero code changes on your end). SDK wrappers for LangChain and LlamaIndex are shipping with the connector release. Provider support expands as the connector release lands — the proxy’s universal design means any new compatible provider works without a code change on rendfly’s side either.

How long are conversations retained?

Default retention is 90 days. You can configure this per workspace down to 7 days or up to 365 days in workspace settings. Conversations older than the configured window are hard-deleted on a nightly job — not archived, not soft-deleted. If you need retention beyond 365 days, that’s an Enterprise conversation; contact us.

Can I bring my own keys (BYOK)?

Yes, on Enterprise tier. With BYOK, judge calls route through your provider account — billed to you directly, with full request logs in your provider’s dashboard. Your provider API key is encrypted at rest (AES-256-GCM) and never logged in plaintext. Indie and Agency tiers use rendfly-managed keys.

Is self-hosting available?

On Enterprise tier. We ship a Helm chart for Kubernetes deployment. Same product, same features, running in your VPC. SOC 2 evidence applies regardless of deployment model — if your auditor needs documentation of the self-hosted deployment, we have it.

What pricing tiers exist?

Indie is $29/month: one workspace, one project, up to 5,000 conversations per month, email and Slack alerts. Agency is $199/month: one workspace, up to 10 projects, 5 users, 50,000 conversations per month, multi-judge consensus, custom rules, all alert channels. Enterprise is custom pricing: unlimited everything, SSO, BYOK, self-hosting, SOC 2 Type II, custom SLAs. Full breakdown at /pricing.

Is rendfly multi-tenant for agencies?

Yes. The Agency tier supports up to 10 projects per workspace, each with its own system message, provider connection, extracted rules, and alert routing. The typical setup for an agency running multiple client agents is one project per client, with a dedicated Slack channel per project so each client’s alerts stay separate. Users on the workspace have access to all projects; per-project access control is on the roadmap for a future release.

Still got questions? Email hi@rendfly.com or read What is rendfly for the bigger picture.