Tools/deepseek-ai/deepseek-harness

DeepSeek Harness

DeepSeek Harness: Everything is a Plugin.

285.3k ★+3.3k/wkestablishedTypeScriptMIT License →

The Lens

By Erik Loyd, SaaS CEO and former COO/CFO of an AWS Premier Partner.

Updated Oct 2026

DeepSeek Harness (dsh) is DeepSeek's own agent app: the layer that gives a model a shell, files, tools, skills and a UI, for coding and everyday work like spreadsheets and slides. Every piece is a Cordis plugin, so you swap parts instead of forking. MIT licensed and free.

npx @deepseek-ai/dsh web starts a local web UI on 127.0.0.1:3080, and DeepSeek now ships an official desktop app with a macOS download on its site. Add a DeepSeek API key, or point it at another provider or any OpenAI-compatible endpoint. Nothing to host.

The harness costs nothing; tokens do. DeepSeek V4-Pro runs $1.32 per million input tokens and $3.96 per million output at weekday peak, half off-peak, and Flash is far cheaper. Built-in web search starts billing November 1. Solo and small teams: free, pin a version. Larger teams: wait.

The catch: it's a developer preview, every release is still tagged pre-release, and DeepSeek's own safety notice says it hasn't had a security audit. August proved it: CVE-2026-82533 (CVSS 9.4) let a sandboxed agent switch off its own sandbox, fixed in 0.1.2-alpha.1. Run it in a VM. The desktop app also sends usage analytics by default, with no switch in the UI.

Free vs Self-Hosted vs Paid

fully free

Free: The whole harness under MIT (copyright DeepSeek): web UI, CLI, desktop app, plugin system, OS sandbox, MCP and skills support. No paid edition and no hosted tier.

Self-hosted: Local only. Node.js plus npx @deepseek-ai/dsh web (serves 127.0.0.1:3080), a pnpm source build, or the Electron desktop app (macOS download on deepseek.com/harness). Desktop reports product analytics by default (device_id, user_id, app version, OS, CPU and memory details, model choice; prompts, responses and API keys excluded); the setting exists in config but has no UI control, and the plain web UI sends none. DeepSeek's safety notice recommends a disposable VM, container or dedicated environment.

Paid: No product to buy; you pay for model tokens. DeepSeek API per 1M tokens at peak (01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday): deepseek-v4-pro $1.32 input (cache miss), $0.044 input (cache hit), $3.96 output; deepseek-flash (now served by DeepSeek-V4.1-Flash-1017) $0.3 input, $1.2 output. Off-peak rates are half, and weekends are entirely off-peak. Search and Fetch APIs cost $0.75 per 1,000 successful requests each; server-side tool calls from the harness hit them and are free until November 1, 2026 (Beijing time). Other providers and custom OpenAI-compatible endpoints also work.

Free and MIT; you pay only for tokens. On DeepSeek's API that's $1.32 input and $3.96 output per million tokens for V4-Pro at weekday peak (half off-peak), plus $0.75 per 1,000 search or fetch calls from November 1.

What to do by team size

Solo
free; run it in a VM or container and stay on current releases
Small team
free; pin a version, the plugin API still breaks between releases
Medium team
free to pilot, but keep it away from real credentials given its security track record
Large team
watch it; pre-release only, no security audit, and desktop analytics you can't switch off from the UI
Self-hosting ops:trivial

Get tools like this every Wednesday

One featured tool, three on the radar. No fluff.

Similar Tools

Score
85/100 · A
Adoption30/30
Maintenance25/25
Community5/20
License15/15
Analysis10/10

A low score is not a verdict on quality. Young and niche tools start low by design. How we calculate scores

About

Owner
deepseek-ai (Organization)
Stars
285,288
Forks
35,475

Explore Further

More tools in the directory

Featured in The Open Source Drop #23