Privacy & data handling

Last updated: 21 May 2026

This page explains exactly what SlyReply does with the data you give it, where it lives, who else sees it, how long we keep it, and the rights you have over it. It is intentionally short and concrete. If something here is unclear or wrong, email privacy@slyreply.ai.

Who is the data controller

The data controller for your personal data is Sherman Studio Ltd — the company that operates the SlyReply service. Sherman Studio Ltd is registered in England and Wales (company number 17244879); its registered office is at 71-75 Shelton Street, Covent Garden, London, WC2H 9JQ. "SlyReply", "we", and "us" on this page mean Sherman Studio Ltd.

For data-protection questions, or to exercise your rights, contact us at privacy@slyreply.ai.

Note: where a customer feeds third parties' personal data through the service (for example, forwarding a customer's email to an agent, or copying an agent into a thread), that customer is the controller of that data and SlyReply acts as their processor — see the Data Processing Agreement.

What we collect

• Your account: name, password hash (bcrypt), registered email addresses, subscription tier.
• Your AI agents (UIDs): inbound address, system prompt, model selection, generation parameters.
• Conversation metadata — for each email thread we keep a small record with no message bodies: the subject line, the participating addresses, message IDs (so replies thread correctly), and timestamps. We never store the contents of the emails themselves or the AI's replies.
• Usage logs (always recorded, no message content): provider, model, tokens used, internal provider cost (USD), timestamp, and which agent handled the message. Used for fair-use accounting, rate limiting, and the spend kill-switch.
• Payment metadata if you subscribe to a paid tier: a Stripe customer / subscription ID. Card details, billing addresses, etc. live with Stripe and never touch our servers.
• Browser-side error reports: when the SlyReply web app catches an uncaught JavaScript exception, it sends us the error message, stack trace (truncated to 10 kB), the URL you were on, your browser's user-agent string, and the app release version. Used to diagnose regressions; visible only in our internal Grafana logs.

Removed in April 2026: we used to store user-supplied AI provider API keys ("BYOK") encrypted at rest. That option has been retired — we now use SlyReply-managed provider keys for everyone, and any old user keys were purged from the database.

Our lawful basis for processing (UK GDPR Art. 6)

Under UK GDPR we must have a lawful basis for each thing we do with your personal data. Ours, per processing activity:

• Running your account and the email pipeline (authenticating you, routing your email to the AI provider, returning the reply, threading conversations) — Art. 6(1)(b), performance of a contract with you.
• Payments and subscriptions (holding your Stripe customer/subscription ID, processing your plan) — Art. 6(1)(b), contract; and Art. 6(1)(c), legal obligation for the tax and accounting records we must keep.
• Fair-use accounting, rate limiting, the spend kill-switch, and abuse / fraud prevention (usage logs, conversation metadata, demo quotas) — Art. 6(1)(f), legitimate interests in keeping the service available, affordable, and free of abuse.
• Security and diagnostics (browser error reports, application logs, the canary monitor) — Art. 6(1)(f), legitimate interests in keeping the service secure and working.
• Optional analytics (Plausible — only when you accept it in the cookie banner) — Art. 6(1)(a), your consent. You can withdraw it any time from the Cookie Policy page.
• Service and security notifications (verification emails, password resets, important account notices) — Art. 6(1)(b), contract.

Where we rely on legitimate interests you have the right to object — see "Your rights" below.

How long we keep your data

Retention periods by category:

• Email and AI-reply contents — not retained. Almost always, an inbound email is processed and its reply sent within seconds, and the contents are never written to our database — they exist only in transit. The single exception is the reprocessing queue (described under "What we don't store" below): when a reply can't be sent right away — your agent is briefly resting under fair-use, or we're retrying past a temporary AI-provider issue — we hold the email just long enough to deliver your reply, and its content is deleted the moment the reply is sent (or if we ultimately can't deliver it). Nothing is kept once your reply has gone out.
• Account data (name, password hash, registered emails, subscription tier, agents/UIDs) — kept for as long as your account is open. When you delete your account, it is removed immediately.
• Conversation metadata (subject line, participating addresses, message IDs, timestamps — no message bodies) — kept while the account is open, then deleted with the account. There is no TTL index and no per-tier retention window: because the rows hold no message content, there is nothing sensitive to age out, and the metadata is needed for threading and troubleshooting for the life of the account.
• Usage logs (provider, model, message size, internal cost, timestamp, agent — no message content) — kept while the account is open for fair-use accounting and abuse investigation, then deleted with the account.
• Payment / billing records — financial records that we are legally required to retain are kept for up to 7 years after the relevant transaction, to meet UK tax and company-law obligations, even after account deletion. Card data itself is never held by us — it lives with Stripe.
• Browser error reports and application logs — short-lived; retained for roughly 30 days in our logging provider for diagnostics, then rotated out.
• Demo (demo@slyreply.ai) metadata — the sender address and per-sender quota counters are kept long enough to enforce the lifetime and daily caps that prevent abuse.

What we don't store

SlyReply does not keep a copy of your conversations. In the normal case the contents of inbound emails and the AI's replies are never written to our database — there is no in-app conversation history, and nothing for us to leak, hand over, or mine. Your own mailbox (inbox + Sent) is the complete record.

One narrow exception, in the interest of full honesty: if a reply can't be generated immediately — your agent is briefly resting under our fair-use cooldown, or we're retrying past a temporary AI-provider outage — we hold your email in a short-lived reprocessing queue so it isn't lost. This is the only place email content touches our database; it lasts only until your reply is delivered (typically minutes to a few hours), and the content is deleted the moment that happens — or if we ultimately can't deliver it. We never keep it beyond that.

Your AI keeps continuity across a thread by reading the quoted message chain your email client adds to each reply (the >-prefixed lines, or "On date, X wrote:" block) — an in-flight pass-through, not stored history. Only the usage logs and conversation metadata described above (no message content) are retained, for fair-use accounting and troubleshooting.

Caveat: not every email client quotes the previous message in replies (some mobile clients trim aggressively, some senders top-post without quotes). When that happens your AI sees only the new message and has no prior context.

Public demo (demo@slyreply.ai)

Anyone can email demo@slyreply.ai without an account to try the service. As with every account, we don't store the contents of demo emails or the replies — only a metadata record (the sender's email address, which we need so the agent can reply to you; the subject; timestamps). Demo use is subject to a per-sender daily quota to prevent abuse. The demo runs on SlyReply's own AI provider keys; no usage is billed back to the sender.

Where it lives

• Database — accounts, agents, conversation metadata, usage logs: MongoDB Atlas, hosted in the EU, encrypted at rest with point-in-time backups enabled.
• Application — API, frontend, inbound SMTP server: hosted in the UK (Contabo), with an EU hot standby (Hetzner).
• HTTP traffic — every request to slyreply.ai transits Cloudflare for DDoS protection, WAF, and (at registration only) bot-challenge via Cloudflare Turnstile.
• Outbound mail — Postmark relays AI replies. They see the envelope (sender, recipient, subject) and message body, and by default retain the message content and deliverability metadata (e.g. bounces) for up to 45 days before deletion; we have not enabled open/click tracking.
• Logs & metrics — Grafana Cloud (Mimir for metrics, Loki for logs). Application logs may include user identifiers (email, account ID) for debugging; message body content is never logged.

We use Plausible for privacy-respecting, cookieless analytics — no personal data, no cross-site tracking, EU-hosted. We do not run advertising trackers or session-replay tools.

Who else sees your data

We share narrow, purpose-bound slices of your data with a small set of processors so the product can function. Their privacy and retention terms apply to the copy they hold:

• AI providers — to generate a reply, the text of the inbound email, together with any supported attachments (PDFs and images), is forwarded to the provider you chose for that agent (Anthropic, OpenAI, Stability AI). This is an in-flight forward for that one request; SlyReply keeps no copy. Anthropic · OpenAI · Stability AI
• Stripe — payment processing for paid tiers. Stripe holds card details, billing addresses, and tax info; we hold only the customer/subscription IDs needed to recognise you. Stripe sets its own fraud-prevention cookies (and may load bot-protection, hCaptcha, on payment pages) — see the Cookie Policy. Stripe privacy
• Cloudflare — HTTP proxy, WAF, and Turnstile captcha at registration. Sees IPs, request headers, and request bodies in transit. Cloudflare privacy
• MongoDB Atlas — managed database hosting in the EU. MongoDB privacy
• Postmark — outbound email delivery. Sees the envelope and body of replies, and by default retains message content for up to 45 days (US-based; UK transfer safeguards apply). Postmark privacy
• Contabo — primary compute hosting in the UK. Contabo privacy
• Hetzner — EU hot-standby compute hosting. Hetzner privacy
• Grafana Cloud — application logs and metrics. Logs may include user identifiers; never message content. Grafana privacy

We do not sell your data. We do not share it with anyone else for marketing, training, or profiling.

International data transfers

Most of our processors are in the UK or the EU. Three of them are based in the United States: the AI providers Anthropic and OpenAI, to which your email text and any supported attachments are forwarded to generate a reply; and the outbound email relay Postmark, which delivers the AI replies. Sending data to any of these means it is transferred outside the UK.

We only make those transfers where there is an appropriate UK GDPR safeguard in place. Depending on the provider, the transfer is covered by:

• the UK International Data Transfer Agreement (IDTA), or
• the UK Addendum to the EU Standard Contractual Clauses (SCCs), or
• a UK adequacy regulation ("data bridge") where one is in force for the recipient — for example the UK Extension to the EU-US Data Privacy Framework, where the provider is certified under it.

You can ask us which mechanism applies to a specific provider, and for a copy of the relevant safeguard, by emailing privacy@slyreply.ai. Remember that SlyReply itself does not store the contents of your emails — the data we forward to the US-based AI providers is an in-flight pass-through for a single request, not a stored copy. The one exception is the outbound relay, Postmark, which (like any email provider) retains the AI reply it delivers for up to 45 days by default before deletion, as described under "Where it lives" and "Who else sees your data" above.

Internal monitoring

We run a small internal canary account that sends a synthetic test email through the pipeline every few minutes so we know the service is working end-to-end. The canary's data is internal-only and is not counted as user data; we mention it for transparency.

How we keep your data safe

The product is small enough that we can list the actual defences end-to-end. Nothing here is theoretical — every item is shipping in the live system today.

• No conversation storage — SlyReply never writes the contents of your emails or the AI's replies to our database. There is no stored conversation history to breach; thread continuity comes from the quoted-reply chain in your own email client.
• In transit — TLS everywhere (Cloudflare + Let's Encrypt). HTTP traffic is forced to HTTPS at the edge; SMTP between us and Postmark is STARTTLS.
• At rest — MongoDB Atlas storage is encrypted; backups inherit the same encryption. Passwords are bcrypt-hashed.
• Auth — sessions are signed JWTs delivered as httpOnly, Secure, SameSite=Lax cookies (no JavaScript can read them). A per-user token version lets us invalidate every session at once on password change or account compromise.
• Inbound email auth — the sender's From address is the authentication for replies; unrecognised senders are silently dropped without revealing whether the address exists. Because that address is the credential, every inbound email is checked against SPF, DKIM, and DMARC-style alignment to defeat spoofing of a registered account: an SPF hard-fail is dropped for everyone, and for Pro/Power accounts a failed DKIM signature or misaligned domain is dropped too (with an operator-managed allowlist for legitimate forwarders). The public demo applies the same alignment checks.
• Free-tier DKIM-fail review — on the Free tier, instead of dropping mail that fails DKIM outright, we record the failure (envelope and authentication-result metadata) so we can investigate abuse and diagnose deliverability problems — for example a misconfigured domain, a shared mailbox, or a forwarding setup that breaks signatures. As with every tier, the message body itself is not stored; only the metadata needed for that review is kept, and only for as long as the review purpose requires. Pro and Power drop DKIM-failing mail immediately and keep nothing. This is also surfaced on your Account settings page.
• Cost-amplification guards — per-sender + per-account rate limits, plus a global daily spend kill-switch. One abusive sender (or compromised account) can't bill the platform into the ground while we sleep.
• Demo abuse defence — disposable-email blocklist (~10k domains), per-IP and per-sending-domain caps, lifetime cap per sender, prompt-injection filter on inbound subjects, output URL filter on AI replies.
• Web application — Cloudflare WAF on every HTTP request; bot-challenge (Turnstile) at registration to keep automated signups out; CSP, X-Frame-Options: DENY, and Permissions-Policy headers shipped on every page.
• Container hygiene — Trivy scans every image we deploy; CI fails on critical vulnerabilities. Production uses pinned image digests, not latest tags.
• Monitoring — every AI call, every billing event, every rate-limit trip emits a metric. Spend, abuse-spike, and login brute-force alerts page Discord. The canary mentioned above tells us within minutes if the inbound pipeline breaks.

Found a gap? Email security@slyreply.ai — we'd rather hear from you than read about it on Twitter.

Your rights (UK GDPR)

If you are in the UK or EU, you have the following rights under UK GDPR. The first three you can exercise from your account directly:

• Right of access / portability: download a JSON export of your data — your profile, your agents, conversation metadata, and any support tickets — from the Account settings page. (A handful of operational records — for example Stripe billing identifiers — are excluded by design, as they are payment-processor references rather than your personal data.)
• Right to erasure: delete your account and all associated data immediately from the same page.
• Right to rectification: edit your profile and email addresses from the same page.
• Right to object / restrict processing: email privacy@slyreply.ai.
• Right to withdraw consent: where we rely on your consent (optional analytics), withdraw it any time from the Cookie Policy page.

Complaints and the supervisory authority

We would always rather you came to us first — email privacy@slyreply.ai and we will do our best to resolve it. But you also have the right to lodge a complaint with a data-protection supervisory authority.

Our supervisory authority is the Information Commissioner's Office (ICO), the UK regulator for data protection. You can contact the ICO at https://ico.org.uk or make a complaint directly at https://ico.org.uk/make-a-complaint/. If you are in the EU you may instead complain to the supervisory authority in your country of residence.

Cookies

SlyReply uses a strictly-necessary session cookie and, only with your consent, privacy-friendly analytics. The full details — and the controls to change your choice — are on the Cookie Policy page.

Changes to this page

When the substance of this page changes, we increment the consent version. Logged-in users are prompted to re-confirm before their next action.