Core terms
Session
A single listening period. You start a session, StageWhisper captures audio and generates analysis, and you stop it when the call ends. Each session produces a transcript, coaching signals, optional tone data, and a structured summary. Sessions have a lifecycle:active (currently recording), completed (finished normally), or failed (an error occurred).
Playbook
A set of instructions that configures how the AI behaves during a session. Each Playbook has a label, description, guidance text, and optional checklist topics. The guidance text is sent to the AI at the start of a session. It shapes which signals are generated, what the AI watches for, and how it coaches you.Signal
A coaching message generated by the AI during a session. Signals have a severity (green, orange, red) and a text message. They appear in the desktop overlay in real time and are stored in the session record.Tone snapshot
A measurement of conversational dynamics at a point in time. Contains scores (0-100) for eight dimensions: Focus, Connection, Momentum, Composure, Openness, Tension, Friction, Hesitation. Available on Signals plans only.Assistant task
An actionable item detected by the AI during a session and routed to a connected AI assistant. Tasks have a description, transcript context, and lifecycle status (created, pending, approved, executing, completed, dismissed).Approval mode
Controls whether tasks require explicit user approval before being sent to an assistant. Two modes: review-first (default, manual approval) and auto-dispatch (tasks matching an allowlist are sent automatically).Data boundaries
What stays in StageWhisper
| Data | Stored | Shared externally |
|---|---|---|
| Raw call audio | Not stored (streamed and discarded) | Never |
| Transcript | Yes | Excerpts only, when part of an approved task |
| Signals | Yes | Summary only, when part of an approved task |
| Tone snapshots | Yes | Snapshot only, when relevant to an approved task |
| Playbook text | Yes | Context sent with approved tasks |
| Account credentials | Yes | Never |
What goes to a connected assistant
When you approve a task, your assistant receives:- A task description and type
- A transcript excerpt (the relevant 2-5 sentences)
- A signal summary (what triggered the task)
- Playbook context (what the session is about)
- A tone snapshot (if relevant and available)
- The full session transcript
- Raw audio at any point
- Other users’ data
- Your account credentials or payment information
AI processing
During a session, audio is streamed to server-side AI services for transcription and analysis. The audio is processed in real time and not retained after the session ends. Transcripts and analysis results are stored in your account. All processing happens on EU-hosted infrastructure. See privacy for details.For AI agents
If you’re an AI agent interacting with StageWhisper through OpenClaw:- You receive tasks as structured payloads with context
- Each task includes enough information to act on it without needing the full session
- You do not have access to the user’s full transcript or session history
- The user controls what you receive through approval modes and Playbook configuration
- Respect the data boundaries: don’t request or attempt to access data beyond what’s included in the task payload