# Security and data

How Spiral handles your API keys, what it stores and for how long, and where your data goes.

## API keys

- A key is a 256-bit random secret, shown once when you create it. Spiral stores only its SHA-256 hash, so a key can never be read back, by you or by us.
- The console shows each key's first 12 characters, so you can tell keys apart.
- Each key has its own spend cap and rate limits. The cap is reserved before a run starts, so it is never overshot.
- Revoking a key takes effect on the next request. A rotated key keeps working for a grace period (24 hours by default), so you can switch without downtime.

## What we store, and for how long

- A run's input, output and timeline can be read in the console and the API for 7 days, so you can check and debug it. After that, only its billing line (tool, units and cost) is shown.
- Billing records, one line per run with its tool, usage and cost, are kept for 25 months.
- Output files and files you upload are stored in Cloudflare R2 for 7 days. Their links are long random names that cannot be guessed from a run, and anyone who holds a link can open the file until it expires.
- Spiral does not use your inputs or outputs to train artificial intelligence or machine learning models.
- Request and response bodies are never written to our logs, and credentials and personal data such as email and IP addresses are scrubbed from them.

The [Privacy Policy](https://spiral.ai/legal/privacy-policy) sets out what personal information we keep and for how long.

## Where your data goes

When you run a tool, Spiral sends your input to the provider that operates it, using Spiral's own account with that provider. You never share a provider API key with us, and the provider sees Spiral, not your API key. Your organization can turn off any provider.

The providers a run can go to: fal, WaveSpeed, Replicate, Hedra, OpenRouter, OpenAI, xAI, Mistral, Gemini, Vertex AI, Anthropic, ElevenLabs, Deepgram, AssemblyAI, Vizard, Exa, Firecrawl, Alexandria, ScrapeCreators and TikHub.

Spiral itself runs on Vercel and Render (the API and the queue worker), Trigger.dev (running tasks), PlanetScale Postgres and Redis Cloud in AWS us-east-1, Cloudflare R2 (files), ClickHouse (run history), BetterStack (logs), Clerk (console sign-in), Autumn and Stripe (balance and payments), Loops (billing email) and Intercom (the support chat on the website).

## Webhooks

Run webhooks follow the Standard Webhooks format: each delivery is signed with HMAC-SHA256 under your organization's secret, in the `webhook-id`, `webhook-timestamp` and `webhook-signature` headers. They go to `https` addresses only and are retried with the same id for up to 24 hours until your endpoint answers 2xx.

## Console sign-in and roles

- Sign in with email, a passkey or your company's single sign-on. An organization can require two-factor authentication for every member.
- SAML or OIDC single sign-on and SCIM user provisioning are set up with your IT admin and us.
- Roles: owner, admin, billing admin, developer and viewer. Owners and admins manage members and every key; only owners, admins and billing admins see the balance and billing.

## Contact

A data processing addendum is available on request. Report a security issue, or ask anything about how we handle data, at support@spiral.ai.

## Common questions

### Security and data

**Does Spiral train AI models on my data?**

No. Spiral does not use your inputs or outputs to train artificial intelligence or machine learning models.

**How long does Spiral keep my inputs and outputs?**

A run's input, output and timeline can be read for 7 days, and output and uploaded files are deleted after 7 days. After that only the run's billing line is kept, for 25 months.

**Does Spiral store my API keys?**

Only a SHA-256 hash of each key. The key itself is shown once, when you create it, and can never be read back.

**Do I need to give you my provider API keys?**

No. Spiral runs every tool on its own account with the provider, so you need one Spiral API key and no provider accounts.

**Which third parties receive my data?**

The provider of the tool you run receives that run's input. Spiral's own infrastructure vendors are listed on this page under "Where your data goes".

**Are output links private?**

An output link is a long random name that cannot be guessed, but anyone who holds it can open the file until it expires after 7 days. Share it only where you would share the file.

**Who in my organization can see runs?**

Every member of the organization can see its runs, inputs and outputs included. Roles decide who manages keys, members and billing.

**Can I get a data processing addendum (DPA)?**

Yes, on request: email support@spiral.ai.

**How do I report a security issue?**

Email support@spiral.ai.

