Integrations

Agent Wispr is intentionally simple: capture speech locally, transcribe it, and insert the result into the application that was focused before recording.

Focused applications

Wispr can support editors, terminals, browsers, chat tools, and other apps that accept normal text input. If an elevated application rejects injection, run Wispr with the same permission level or use the history panel to copy the text.

Local model settings

Choose the installed model in Settings. Smaller models reduce latency and resource use; larger models can improve accuracy. The initial model download is separate from ordinary offline use.

History and export

Transcriptions remain available in the local history panel. Search, re-inject, export, or delete entries from the UI. Remove private text before sharing an export or support screenshot.

The public docs do not publish private endpoints, credentials, or internal implementation details.