AI tokens

Hyprcore Cloud measures AI usage in tokens. They cover the AI work that runs on Hyprcore-managed infrastructure: meeting summaries, action items, knowledge base and session chat, and LLM post-processing of your dictation and transcripts.

Plan	Monthly token budget
Free	None (no Hyprcore Cloud AI)
Pro	5,000,000 tokens / month
Team	12,500,000 tokens / month (shared across the team)

Your monthly budget doesn't roll over—it resets at the start of each billing cycle. Purchased token packs are separate: they never expire, and they're only drawn down after your monthly allowance runs out.

Note: The Free plan has no Hyprcore Cloud AI budget. You can still run local models on your Mac, or add your own API keys (BYOK), to use AI features for free.

What consumes tokens

Anything that calls a Hyprcore-managed cloud model spends tokens. In practice that's:

• AI meeting summaries — generating the summary of a recorded meeting.

• Action items — extracting tasks and follow-ups from a meeting.

• Knowledge base and session chat — asking questions across one meeting or your whole knowledge base.

• LLM post-processing — cleaning up and polishing dictation and transcripts.

Longer meetings, larger context, and smarter models use more tokens than short, single-meeting requests. Work that runs on a local model or through your own API key (BYOK) costs zero tokens.

See your usage

app.hyprcore.ai/dashboard/usage shows:

• AI Tokens Used this cycle, against your total available budget

• Tokens remaining

• A breakdown by service—Language Model, Transcription, and Embeddings

• Recent activity, request by request

• For Team plans: usage is drawn from the shared pool

If you've bought token packs, the dashboard shows your purchased top-up balance folded into the total. The desktop app's Account settings shows the same totals at a glance.

Stretch your tokens

• Use local models for routine work. Local Whisper / Parakeet for transcription, Ollama (or another local model) for cleanup. Free, and nothing leaves your Mac.

• Pick smaller LLMs for cleanup. Lighter models handle short transcripts well, often without a noticeable quality drop.

• Bring your own keys. Plug in your Anthropic, OpenAI, Deepgram, or Groq keys. Hyprcore makes the call as you, using your key, and your provider invoices you directly—no tokens spent.

• Skip cloud transcription for short meetings. Local Whisper is fine under 30 minutes.

Run out of tokens

When your monthly budget hits zero:

• AI summaries, action items, and chat stop unless you've configured a BYO key as a fallback.

• Cloud transcription falls back to local models where your provider policy allows.

• Sync keeps working—it doesn't use tokens.

• Local features keep working—dictation, local transcription, vault, and search.

You can:

• Wait for the cycle to reset.

• Buy a one-time token pack from Billing → Buy tokens. Packs never expire.

• Configure a BYO key in Settings → Models → Post-processing so Hyprcore can route around token exhaustion.

Bring your own keys

Adding a BYO key removes the token cost for that provider entirely. Hyprcore makes the API call as you, using your key. Your provider invoices you directly.

To add a key, see the relevant model setup page:

• Speech models (Deepgram, Groq, OpenAI Whisper)

• Post-processing LLMs (Anthropic, OpenAI, Google, Groq, OpenRouter)