Everything we get asked about security, pricing, how HermesRecall actually runs, and how to reach us.
No — not during support, not for debugging, not ever. Your Honcho memory entries and chat logs are encrypted at rest with per-tenant keys that live only inside your private instance. Our operators can restart your container, scale compute, and ship updates, but they cannot decrypt or read a single message. When you open a support ticket, you choose what to share — we never pull data from your instance without your explicit consent in that ticket.
Data in transit is protected by TLS 1.3. Data at rest is encrypted with AES-256-GCM. API keys are encrypted with a separate KMS-managed key. Each subscriber runs in their own isolated instance with its own storage volume — nothing is shared with other users.
Your agent and memory keep running until the end of the billing period you already paid for. At the end of that period, we tear down your private instance, wipe the Honcho memory layer, and permanently delete all your conversation data within 24 hours. You can also export everything beforehand from the dashboard — conversations, memory entries, settings — as a portable JSON archive.
No. You don't need a server, an API key, a Fly account, a Python version, or a weekend. We handle the infrastructure, the AI provider, the skills, the updates, and the monitoring. You need a phone and a Telegram account. That's it. If you later want to bring your own AI provider key or install a community skill, that's available in the dashboard — but it's opt-in, not required.
Three steps, under two minutes end-to-end. First, we confirm your plan and spin up your dedicated instance. Second, you link Telegram — scan a QR code. Third, we run a handshake to make sure the bot responds, then drop you into your dashboard with your first conversation ready. You can optionally plug in your own AI provider key during step one if you prefer BYOK — most subscribers skip it and use the managed provider we ship with.
Yes — no lock-in, no penalty, no "call us to cancel" trick. Cancellation is one click in your dashboard. You keep full access until the end of the billing period you already paid for.
No refunds for partial periods, but you keep access through the end of the period you paid for. If we miss our 99.5% monthly uptime target, you get a pro-rated credit on your next invoice — request it at support@hermesrecall.com.
Yes. Upgrades take effect immediately and are pro-rated for the current period. Downgrades take effect at the next billing cycle so you don't lose access to paid features mid-period. Both are one click from the Plan & Billing tab.
Not right now. We run dedicated infrastructure for every subscriber, which makes a free tier expensive to keep honest. The three paid tiers start at $29/mo and are cancellable anytime with no commitment.
Yes — every HermesRecall plan runs on a dedicated instance, not a shared cluster. Your agent, your memory, your data, your server. Nothing is shared with other users. This means consistent performance (no noisy-neighbor slowdowns), full isolation for privacy, and the ability to restart or scale independently.
No. Not a shared container, not a multi-tenant worker, not a pool. Your agent has its own compute, its own memory layer, its own encrypted storage, and its own network identity on Fly.io. That means cryptographic isolation at the storage layer and consistent latency regardless of what anyone else is doing. You can restart your instance without touching anyone else's.
Instances run on Fly.io's global network. By default we provision in the region closest to your signup IP to minimize latency. If you need to move your instance to another region (for residency reasons or latency), open a support ticket — we handle the migration manually and preserve your memory and conversation history during the cutover.
Sixty seconds from payment confirmation to a live agent, in the median case. We provision the microVM, boot the Hermes runtime, attach the Honcho memory layer, and pair Telegram — all scripted end-to-end. If provisioning exceeds five minutes we page our oncall automatically.
Zero touch on your side. We handle every update: the Hermes Agent core, the Honcho memory layer, all curated skills, security patches, dependency upgrades — everything. You always run the latest stable version. Updates roll out during low-traffic windows with zero downtime.
Hermes Agent is a mature open-source AI assistant framework — fast-moving, used by thousands of developers. It's missing one thing: memory across sessions. HermesRecall is an independent managed service that runs Hermes Agent for you, on dedicated infrastructure, with the persistent memory layer it never had. We are not affiliated with the open-source project.
Every plan includes Telegram and the web dashboard chat. WhatsApp goes live April 26 for Max subscribers — each gets a dedicated WhatsApp Business number. Memory is shared across all channels.
Yes. You can install community skills from our curated marketplace, and on Pro and Max you can write your own skill in TypeScript or Python and upload it to your private instance. Custom skills stay private to your account. Marketplace skills are vetted by us before listing.
Open the Support tab in your dashboard and send a ticket — or email support@hermesrecall.com. We aim for a first response within one business day on Lite and Pro, and within four hours on Max.