MISTRAL VOICE TOOL SURFACE WORKSHEET Free original planning resource | Reviewed October 24, 2026 Companion: https://burki.dev/blog/mistral-voice-agent-browser-phone-tools No account, agent, session, connector, SIP route or provider call was created for this resource. Fixtures are fictional and unperformed. Provider and production services can have separate costs. 1. TASK AND APPROVED INFORMATION Task: What the assistant may explain: Actions outside this version: Fact source / record ID / version / owner: Approved next step if information is unavailable: Fictional record: CRAFT-SAT / entry-01 Saturday craft workshop: use the garden entrance. No reservation, admission, attendee data or actual workshop offer is supplied. 2. TOOL INVENTORY, ONE ROW PER DEPENDENCY Tool name and expected inputs: Declaration stored in: Actual handler or approved source: Who owns and authorizes that component: Execution surface: browser / local SDK / provider phone session Evidence that the component is available there: Read-only result / allowed business effect: Unresolved dependency: 3. SURFACE PLAN Browser application release: Authenticated server route creating sessions: Provider secret reference (never the key value): Approved options the server accepts: Client handler attached for the intended tool: Session close path and closure evidence: Phone eligibility / provider account-team confirmation: Saved agent and version intended for phone use: Authorized server-side source actually available: Required connector authentication / document readiness: Client-only dependencies removed or explicitly unavailable: Supported next step when a human is needed: Do not assume a browser handler runs inside a provider phone session. Do not relabel an arbitrary client function as a server-side connector. 4. ORIGINAL EXPECTED CASES, ALL UNTESTED A. CRAFT-SAT and approved record available. Expected: explain garden entrance, without promising admission. Observed result / evidence / correction: B. Unknown workshop code. Expected: missing fact remains unknown; no answer borrowed from CRAFT-SAT. Observed result / evidence / correction: C. Declaration exists but the session has no handler. Expected: retain failure evidence; no successful lookup claimed. Observed result / evidence / correction: D. Browser-only function appears in a proposed phone configuration. Expected: stop the deployment assumption; identify an available authorized server-side source or narrow the task. No invented migration switch. Observed result / evidence / correction: E. Approved document or connector unavailable. Expected: explain the factual limit and the operator-approved next step. Observed result / evidence / correction: F. Caller requests a person or a callback. Expected: describe the actual supported process; do not claim an unsupported transfer or delivered callback. Observed result / evidence / correction: G. Short waiting message plays, but the lookup fails. Expected: waiting speech does not become success evidence. Observed result / evidence / correction: 5. COST AND ACCEPTANCE Provider product and applicable tariff / date: Conversation minutes or other exact units: Other carrier, number, platform, connector and support costs: Configuration evidence: Provider-accepted session evidence: Tool result evidence: Caller-heard answer evidence, if authorized and applicable: Actual business outcome / unknowns: Decision owner / next review trigger: CURRENT CONTRACTS, RECHECK BEFORE IMPLEMENTATION https://docs.mistral.ai/studio/audio/voice-agents/tools https://docs.mistral.ai/studio/audio/voice-agents/browser https://docs.mistral.ai/studio/audio/voice-agents/telephony https://docs.mistral.ai/inference/pricing