DeepSeek Harness vs Kimi Code
Comparer des modèles
DeepSeek Harness is more worth it — 79% vs 39%
How we score worth-it? →Open-source agent framework — everything is a plugin
- Logs that aren't missing anything out of the box. Pretty underused concept in the time of 8TB SSD drives.— hn
- The other thing that matters to me is the tooling. I could use Deepseek Harness with Claude but I'd rather support the same people that brought me DSH.— hn ↗
- DeepSeek harness seemed really powerful when I tried it.— hn ↗
- I have been using deepseek harness web app atop my local model and am already liking the better organization and UI of this interface. AND it is all open source/weights.— hn ↗
- Switched to DeepSeek Harness recently, their new Creator Mode is amazing.— hn ↗
- I wouldn't recommend beginners go messing around with DeepSeek Harness right now.— bilibili ↗
- DeepSeek Harness — I feel most of its hype comes from non-functional things. What people play with most in plugins is the UI (other products' extension mechanisms generally can't customize the UI this heavily) and whale-girl content. Nothing else seems to be an essential, irreplaceable difference.— v2ex ↗
- Personally I feel dsh right now is just the dumplings made to go with the cordis vinegar — the UX shows no obvious improvement over Codex or cc (subjective experience).— v2ex ↗
- If it didn't ship with a built-in API it'd be fine; with one built in it feels like the kind of software that scrapes my keys and projects.— bilibili ↗
“Open-source agent framework where everything is a plugin.”
“Plugins, skills, MCP — you can wire up anything.”
“Great for self-hosted agent workflows and tinkerers.”
Kimi's coding agent
- I also like Reasonix quite a bit, as an agent harness, though Kimi Code is also very good.— hn
- Claude (top tier), GLM top tier, Kimi top tier... I'm using Codex, Claude Code, Kimi Code for different things... I max out the Claude every week and the Kimi most weeks.— hn ↗
- Kimi is my daily driver. You folks are doing amazing work. I wish you guys all the best and I look forward to 2.7 hopefully we get some agentic usage upgrades.— youtube ↗
- Subagent model selection is not honored after switching models... The runtime continues using the originally selected subagent model.— github ↗
- I believe my monthly quota is being consumed incorrectly. Over roughly 2 weeks I used ~88% of my monthly quota while doing very little actual work — mostly small chat queries and a couple of short sessions.— github ↗
- the LLM request contains zero `mcp__*` entries, so the agent cannot call any MCP tool at all.— github ↗
- Whatever you do, do not subscribe to Kimi - it is so slow you start doubting life; basically you pay money to be unable to use it.— bilibili ↗
- Turns that kimi starts on its own from background-task notifications are never forwarded over ACP (the host sees nothing)— github ↗
“Fast, cheap, and the CLI is clean”
“Good for Chinese developers who want something local-ish”
“Less mature than the top agents, but improving fast”
Which should you pick?
The gap is clear: DeepSeek Harness sits at 79% while Kimi Code is at 39%, a 40-point difference. If you just want the safer pick, start with DeepSeek Harness.
They price differently (DeepSeek Harness: Free (open-source); Kimi Code: $15–$159/mo), so estimate by your own usage.
Both are in the Agent category with similar coverage; the real difference is stability and cost — run one of your own tasks through each before deciding.
What users actually said about DeepSeek Harness: "Logs that aren't missing anything out of the box. Pretty underused concept in the time of …"
FAQ
Is DeepSeek Harness or Kimi Code more worth it?
DeepSeek Harness is more worth it, with a 79% worth-it score vs 39% for Kimi Code.
How much does DeepSeek Harness cost vs Kimi Code?
DeepSeek Harness: Free (open-source) · Kimi Code: $15–$159/mo.
What do users say about DeepSeek Harness vs Kimi Code?
DeepSeek Harness: Open-source agent framework where everything is a plugin.. Kimi Code: Fast, cheap, and the CLI is clean.