What problem does Agent Terminal solve?

It bridges voice input, semantic memory, and AI task delegation into one desktop terminal so you can speak a goal and have it routed to the right tool without leaving your workflow.

Configuration

Agent Terminal is configured via environment variables. No config file is required for the default setup.

Core variables

Variable	Purpose	Default
`BRAIN_BASE_URL`	Agent Brain base URL	unset (disabled)
`HARNESS_BASE_URL`	Agent Harness base URL	unset (disabled)
`TERMINAL_DB_PATH`	SQLite history database path	`~/.agent-terminal/history.db`
`HARNESS_TOKEN`	Bearer token for Harness auth	unset

Pre-send guard

The guard is implemented in src/preSendGuard.ts and evaluated through POST /api/guard/evaluate. It asks Brain for reusable candidates and rewrite suggestions, then returns one of reuse_answer, rewrite_prompt, send_original, or surface_chooser. It does not silently block a provider request, and it fails open when Brain is unavailable.

Wispr integration

To enable voice transcript ingestion from Agent Wispr:

Enable the Terminal integration toggle in Agent Wispr Settings
The Wispr app can POST transcripts to POST /api/wispr/ingest on the terminal backend
Transcripts appear in the Transcripts tab for review before sending

◆Guard is immutable at runtime

The pre-send guard cannot be disabled or bypassed at runtime. Modify src/preSendGuard.ts in source and rebuild to change guard behavior.