The deciding difference is operational: DeepSeek Harness can run on infrastructure you control, Claude Agent SDK cannot. If that constraint is real for you, it settles the choice before anything else is considered.
The Claude Code loop as a library, for agents that are not about code.
Same harness as Claude Code, but you own the entrypoint and the tool surface.
Everything is a plugin — the model adapter, the tool registry, the sandbox and the agent loop itself are all swappable.
Web UI first — `dsh web` serves on 127.0.0.1:3080 rather than taking over the terminal. Built on Cordis. Still a developer preview: DeepSeek warns of compatibility-breaking changes, so pin the version.
| Claude Agent SDK | DeepSeek Harness | |
|---|---|---|
| Licence | Proprietary | Open source |
| Pricing | usage based | open source |
| Self-hostable | No | Yes |
| Languages | TypeScript, Python | TypeScript |
| Install | npm install @anthropic-ai/claude-agent-sdk | npx @deepseek-ai/dsh web |
| Keys required | ANTHROPIC_API_KEY | None |
The deciding difference is operational: DeepSeek Harness can run on infrastructure you control, Claude Agent SDK cannot. If that constraint is real for you, it settles the choice before anything else is considered. Claude Agent SDK: The Claude Code loop as a library, for agents that are not about code. DeepSeek Harness: Everything is a plugin — the model adapter, the tool registry, the sandbox and the agent loop itself are all swappable.
Claude Agent SDK is a hosted service only. DeepSeek Harness can run on your own infrastructure.
Claude Agent SDK is proprietary (usage based). DeepSeek Harness is open source (open source).
Every page here answers to Accept: text/markdown and returns the same content at roughly a tenth the tokens. No separate site, no toggle — same URL.
curl -s -H "Accept: text/markdown" https://newagent.build/compare/claude-agent-sdk-vs-deepseek-harness