DSH Baton

•
244
•

Delegate coding tasks from AI assistants like Claude Code and Codex to a local DeepSeek Harness instance with live progress tracking and automated result validation.

DSH Baton acts as an orchestration bridge that hands off intensive coding execution from primary AI assistants such as Claude Code, Codex, and Antigravity to a local DeepSeek Harness environment. Tasks run interactively and visibly within the local Harness window, allowing developers to observe changes or intervene at any time. When execution finishes, DSH Baton reports back directly to the primary agent without polling, delivering a comprehensive summary of modified files, automated test outcomes, and requirement checklists while significantly reducing rate limits and token costs.

主な機能

01Live task monitoring in the DeepSeek Harness interface with interactive mid-run steering
02Substantial quota and token conservation on primary models like Opus and GPT-6
03Zero-polling result handoff returning completion status within seconds
04Automated evidence collection including modified files, native test runs, and requirement checklists
05Clear multi-agent message provenance tracking showing originating agents
06240 GitHub stars

ユースケース

01Running multi-phase full-stack feature development with automated test verification between iterations
02Offloading heavy, repetitive implementation tasks to reduce usage against premium model rate limits
03Separating architecture planning in Claude Code from execution and testing in DeepSeek Harness