Agent workflows

Let AI agents transcribe local files directly

After installing the EdgeSpeak Skill, agents can call edgespeak-cli for local audio and video.

Direct answer

The Skill teaches agents how to call the CLI. EdgeSpeak transcribes on-device and returns the result to the task.

Install the Agent Skill

The default command detects the current agent. Use the full install guide when you need an explicit --agent target.

npx skills add lattifai/EdgeSpeak-Skills

Start transcribing

After installation, the agent calls the same CLI. You can also run it directly:

edgespeak-cli transcribe <audio-or-video-file>

Two runtime modes

When the app is running, the CLI uses the local gateway and warm model. Without the app, the CLI starts its bundled on-device engine.

Agent transcription paths

DimensionEdgeSpeakOther workflow
InstallAgent Skill + edgespeak-cliCustom hosted API wiring
AudioStays localUsually uploaded
Best forDesktop agents and local automationServer agents and batch jobs

FAQ

Which agents are supported?

Any Agent Skills client can install it. Official examples include Claude Code, Cursor, and Codex.

Must the desktop app be open?

No. Without the app, the CLI can start its bundled on-device engine.

Is this the same as MCP?

No. The Skill teaches agents to call the CLI. MCP is a separate integration path.

Evidence links