BYOK + dedicated runtime
Your own AI, your own key, your own server — without the ops work.
Self-hosting AI traditionally means a VPS, Docker, fragile updates, and 2am downtime. Hermify Dedicated is a managed self-hosted experience: you bring your own AI key, we provision a dedicated server, and you get memory, Telegram, and custom skills out of the box.
You want full control over the AI provider and pay for actual usage, not a flat subscription.
You want isolation — no shared infrastructure with other users.
But you do not want to babysit a server, manage Docker, or rotate secrets manually.
BYOK ("bring your own key") means you keep your relationship with OpenAI, Anthropic, OpenRouter, or any compatible provider. You pay them directly for usage, and Hermify never sees your AI bill.
Hermify's job is the runtime: memory, Telegram, dashboard, skills, and a stable container that runs your assistant 24/7. Your AI provider's job is the model. Clean separation, no markups.
On the Dedicated plan, Hermify provisions a private server bound to your account. Your assistant runs in its own container with its own resources.
No noisy-neighbor issues, no shared memory, no shared model context. Just your AI, your key, and your server.
Coming from ChatGPT? Sign up, paste your OpenAI key, connect Telegram, and you are running. Memory will start filling in immediately.
Coming from a hand-rolled self-hosted setup? Cancel your VPS, point your existing API key into Hermify, and your assistant is back online in minutes — with memory, dashboard, and Telegram included.
Starter
$19/mo
BYOK on shared infrastructure (still per-user isolated container).
Pro
$29/mo
Managed AI access included if you want to skip provider setup.
Dedicated
$49/mo
Your own private server with BYOK or included AI access. Maximum isolation.
You skip the VPS rental, Docker setup, secret management, restart scripts, and 2am downtime. You keep BYOK, isolation, and full control of the AI provider.
Yes. Any OpenAI-compatible endpoint works. OpenRouter is the most flexible (one key, hundreds of models). Local models work too via a compatible endpoint.
Dedicated gives you your own server. Starter shares infrastructure (with isolated containers per user). Dedicated is right when you want maximum isolation, more headroom, or just feel safer with private hardware.
No. With BYOK on Starter or Dedicated, prompts go directly from your runtime to your AI provider. We never read or log them.
Yes. Memory and skills are exportable from the dashboard at any time.
Bring your own OpenAI, Anthropic, or OpenRouter key. Pay actual usage. Get memory, Telegram, and custom skills for $19/mo.
Open page ›Compare Hermify and ChatGPT Plus. Persistent memory, Telegram access, custom skills, and bring-your-own-key turn AI from a chat tab into a real personal assistant.
Open page ›Compare managed Hermes Agent hosting with self-hosting on a VPS. See when Hermify is the faster, lower-friction path to production.
Open page ›