@Saccc_c: Deepseek harness's trace mode is so interesting, now you can completely know the progress of Agent's work, and your Agent employees will never dare to slack off again. Its specific thinking steps, tool calls, and tool return results are all clearly visible. The most interesting thing is that it uses different colors to distinguish when AI executes tasks…
Summary
DeepSeek Harness has been released, featuring a trace mode that visualizes Agent's thinking steps, tool calls, and return results, and uses colors to distinguish different types of content. It can be installed via npm and is suitable for developers to observe Agent work progress.
View Cached Full Text
Cached at: 08/14/26, 03:33 AM
Deepseek harness’s trajectory mode is so fascinating—now you can see exactly what the Agent is working on at every step. Your Agent employees won’t dare slack off anymore.
You can clearly see its specific reasoning steps, tool calls, and tool return results. The most interesting part is that it uses different colors to distinguish various content generated while the AI executes tasks—for example, orange represents tool calls, green represents context, and it displays them at the top like a trajectory.
Sac (@Saccc_c): DeepSeek Harness is here!
Just run these two commands in your terminal to install and launch it:
1、npm i @deepseek-ai/dsh
2、npx @deepseek-ai/dsh web
Coding beginners, stay away! The four built-in modes are quite confusing—gonna need some time to figure them out 🧐
Similar Articles
@Saccc_c: In Deepseek Harness's trajectory mode, you can also view its system prompt (see Figure 1). I translated it into Chinese; interested friends can take a look (see Figures 2 and 3)
The tweet introduces Deepseek Harness's trajectory mode, which allows viewing the Agent's thinking steps, tool calls, and system prompts, accompanied by a Chinese translation for easy learning.
@seekjourney: Everyone may have missed a hidden treasure in DeepSeek Harness's release yesterday: .agents/skills/ directly contains 11 real engineering Skills: Code Review, Pre-push Checks, finding over-engineering, Stacked…
Introduces the .agents/skills/ directory included in the DeepSeek Harness release, containing 11 real engineering Skills (such as code review, pre-push checks, CoT leakage cleanup, etc.), which is equivalent to open-sourcing part of their engineering standards and Agent SOP.
@jakevin7: https://x.com/jakevin7/status/2086031167040426488
Maka is an open-source Agent Harness. Through mechanisms such as log-as-runtime, context pruning, and thinking feedback, it cuts the cost of the same DeepSeek task to 1/8 of OpenCode, while achieving a higher pass rate on Terminal-Bench at lower cost.
@wsl8297: For those who usually use DeepSeek for coding, check out DeepSeek-Code-Whale. GitHub: https://github.com/usewhale/DeepSeek-Code-Whale... Open-source terminal AI coding agent, specialized...
DeepSeek-Code-Whale is an open-source terminal AI coding agent, specifically optimized for DeepSeek models, supporting MCP tools, Skills extensions, prefix caching optimization (90% cache hit rate) and 1M context window, aimed at reducing AI coding costs and providing efficient command-line workflows.
@BoxMrChen: The Deepseek Harness philosophy is awesome. They made everything into plugins — GUI, TUI — and oppose hardcoding workflows like plan, subagent, permissions, MCP into the core. To be honest, at first glance I felt they reinvented Pi Agent, even though it was built on Cordis...
The author comments on Deepseek Harness's plugin-based philosophy, arguing that it turns everything (GUI/TUI, etc.) into plugins and opposes hardcoding workflows. They compare it with Pi Agent and Prime-Agent, propose the idea of building a self-evolving Agent in a REPL, and have already started implementing it with Codex.