# Capabilities

> Source: https://whisperer.one/llms/features.md · Whisperer machine-readable layer.
> Product overview: https://whisperer.one/llms/product.md · Full corpus: https://whisperer.one/llms-full.txt

The full capability set, generated from the product's capability registry — the same source
the site's `SoftwareApplication.featureList` and the bullet list in `llms.txt` are built from,
so the three cannot disagree.

| Capability | What it does | Documentation |
| --- | --- | --- |
| `recording` | Listens: live transcription of microphone and system audio, with speaker labels and timestamps. | /docs/en/features/live-transcription.html |
| `realtime-answers` | Suggests: real-time answers in a desktop overlay that stays invisible in screen sharing; every suggestion is built from the running transcript, your knowledge base and your profile, not from one isolated question. | /docs/en/features/ai-suggestions.html |
| `knowledge-base` | Searches your own knowledge: notes with wiki-links and a knowledge graph; retrieval-augmented generation (RAG) mixes them into every answer. | /docs/en/concepts/knowledge-base.html |
| `screen-vision` | Sees: analyses screenshots and what is on screen — diagrams, charts, slides, code, terminal errors. | /docs/en/features/screenshots-and-vision.html |
| `translation` | Translates: live translation — the other side speaks their language, you read yours as the call runs. | /docs/en/features/realtime-translation.html |
| `model-choice` | Adapts: a model chosen per role and per task across OpenAI, Anthropic, Google, DeepSeek, Qwen, Moonshot, xAI and NVIDIA, on one subscription. | /docs/en/concepts/model-catalog.html |
| `custom-prompts` | Runs your own prompts: you write the prompt, attach skills, pick the model and the context each answer role uses. | /docs/en/features/prompt-studio.html |
| `ai-notes` | Documents: turns the transcript into structured notes and action items. | /docs/en/features/history-and-export.html |
| `after-meeting` | Summarises: a mind map of decisions and analytics after the call, with export to Markdown, SRT, VTT and JSON. | /docs/en/features/mindmap-and-analytics.html |
| `calendar-rooms` | Calendar and video rooms: a room lives inside the event, guests join by link; two-way sync with Google Calendar, Outlook and CalDAV. | — |
| `tracker` | Tasks: decisions from a call become cards on a kanban board with due dates, priorities and labels. | /docs/en/features/tracker.html |
| `files` | Files: folders, search and drag-and-drop upload; room recordings land there and become a meeting with a transcript. | — |
| `integrations-mcp` | Connects: Notion, Slack, HubSpot and Zapier in your account, plus an MCP server that lets Claude, ChatGPT and Cursor read and act on your Whisperer data. | /docs/en/integrations/mcp.html |
| `agent-leo` | Acts: Leo, the built-in AI agent, sees your meetings, tasks, calendar and files and acts in the same places. | /docs/en/features/leo-agent.html |
| `coding-role` | Writes code: a dedicated coding role that runs on its own model, chosen independently of the rest. | /docs/en/concepts/model-roles.html |
| `system-design` | Works through system design: a structured answer format with component diagrams and the reasoning behind each decision. | /docs/en/features/system-design-mode.html |
| `any-call` | Works on any call: Zoom, Google Meet, Microsoft Teams, Telegram, Discord and others — without bots, plugins or per-service integrations. | /docs/en/integrations/supported-services.html |

## Notes on reading this table

- A capability with no documentation link is real and shipped; it simply has no dedicated
  Help Center page yet.
- The first thirteen rows are the ones shown as cards on the product site. The rest are
  shipped capabilities that are not marketed as separate cards — the agent, the coding role,
  the System Design answer format and the fact that the product works on any call.
- Plans gate some capabilities: the knowledge base requires a paid plan, and frontier models
  require Max. Multi-screenshot vision and System Design mode are not plan-gated — they work
  on Free too. See `/llms/pricing.md`.