DeepSeek

DeepSeek runs through the DeepSeek Harness CLI, dsh. Install it with npm install -g @deepseek-ai/dsh, then sign in from Settings → Harness and pick DeepSeek as a project's worker or orchestrator agent, or as the default harness. The harness name for the CLI is deepseek-harness.

tallpine agent ls --refresh
tallpine spawn --project my-app --name ds-task --agent deepseek-harness --mode chat

Interfaces

• Terminal UI: a one-shot task. Tallpine runs dsh --profile headless "<task>", which answers the task, prints the result and exits. The task must be non-empty. Use Chat for an ongoing conversation.
• Chat: Tallpine binds to dsh's acp profile, an Agent Client Protocol server. Session activity comes from this Chat stream, because dsh has no workspace hook file for Tallpine to install.

Models and reasoning effort

The terminal profile exposes no model or reasoning-effort flag, so it uses the profile's own configured route. Per-session model and reasoning-effort selection are available in Chat only.

Privacy

Every shipped dsh profile mounts a session-log plugin that uploads the session log, including prompts and tool output, with official-API requests. Tallpine launches dsh with a profile overlay that turns that upload off, so session logs stay local.

Limits

DeepSeek is a worker and orchestrator harness only. It is not offered as a code reviewer; reviewers are Claude Code, Codex and Kiro.