MIT-licensed Python search + evidence-verification toolkit for agents, shipped as a DeepSeek Harness plugin bundle (10 MCP tools, optional wide_research). Zero-config on free/local engines, multi-language routing, dual-layer caching, and a closed-loop evidence gate for high-stakes queries.
DSH integration
Ecosystem-related
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-25
License
MIT
01What can it help you accomplish?
Use Argo as the search backend for a DeepSeek Harness agent
Compact 'evidence candidate + credibility breakdown' JSON (selection / absorption / credibility_fast / evidence_flags) the agent can sort and fetch-verify
DeepSeek Harness agents and retrieval workflows that need grounded, checkable evidence
Run deep research on a broad survey / investigation question
Sub-questions decomposed, multi-source parallel collection, and a 'what evidence is still missing' gap report (dossier)
Agents doing literature/research surveys, fact-checking, and multi-source cross-checks
Verify evidence for high-stakes queries (finance / medical / legal)
High-stakes results flagged fetch_required, each result flagged fetch_suggested; `--verify` fetches and back-fills a 'post-verification evidence score', and verified links are remembered
Agents doing fact-checking, public financial info, and entity multi-source cross-checks
02How to install into DeepSeek Harness
Prerequisites
- Python 3.10+ is required for the CLI and MCP core; the only hard dependency is `pip install pyyaml`.
- Node.js 18+ is needed only if you use the `npx` entry point (the DSH plugin path is pure Python).
- No API key is required — free engines and local `local_*` engines work out of the box, and keyed sources are auto-skipped when unset.
Installation steps
- 01
Install the Argo DSH plugin bundle (10 `mcp__argo__*` MCP tools): `dsh plugin --profile web add "github:taxueseek/argo"`
$ dsh plugin --profile web add "github:taxueseek/argo"
- 02
Optionally add the wide_research parallel-research subpackage: `dsh plugin --profile web add "github:taxueseek/argo#main&path:packages/dsh-plugin"`, then restart `dsh web`.
$ dsh plugin --profile web add "github:taxueseek/argo#main&path:packages/dsh-plugin"
Verify the integration
- python3 ~/.local/share/argo/scripts/search.py "贵州茅台股价" --json
- python3 ~/.local/share/argo/scripts/search.py --list-engines
03DSH integration and capability boundaries
Installs as a first-party `.dsh-plugin` bundle via `dsh plugin add github:taxueseek/argo`, exposing 10 `mcp__argo__*` MCP search tools to DeepSeek Harness (an optional subpackage adds `wide_research` parallel research orchestration); it connects through the MCP layer of the DSH web profile.
Unified + vertical search
Any natural-language query (Chinese / English / Japanese / Korean, and more)→Routes to the right vertical source (quotes, film, sports, macro, academic…) and returns compact evidence JSON — ~120+ engines, 60+ domains
Deep research orchestration
A broad survey / investigation question→Decomposes into sub-questions, parallel multi-source collection, gap report
Evidence verification closed loop
High-stakes queries (finance / medical / legal)→fetch_required / fetch_suggested flags + `--verify` back-fills post-verification evidence score; verified links remembered
10 MCP tools (incl. local file search)
MCP calls from the DeepSeek Harness agent→Unified JSON via 10 tools: argo_search / argo_local_search / argo_research / argo_evidence / argo_clarify / argo_fetch / argo_crawl / argo_screenshot / argo_pdf / argo_social_search
04Who is it for? When not to use it?
Good for
- DeepSeek Harness agents and retrieval workflows that need grounded, checkable evidence
- Agents doing literature/research surveys, fact-checking, and multi-source cross-checks
- Agents doing fact-checking, public financial info, and entity multi-source cross-checks
Not for
- Do NOT run `npm install argo-search` — the npm registry copy is an unofficial, stale v1.0.1 not maintained by this repo (incomplete features, not updated with this project). This package sets `private: true` in package.json to prevent accidental publishing.
- Login-state professional search (ego-search) is off by default and needs a real browser via ego lite (macOS only) or the WebBridge extension; it is not available out of the box and requires extra setup.
05Compatibility, maintenance and safety notes
- Do NOT run `npm install argo-search` — the npm registry copy is an unofficial, stale v1.0.1 not maintained by this repo (incomplete features, not updated with this project). This package sets `private: true` in package.json to prevent accidental publishing.
- Requires Python 3.10+ for the CLI and MCP core (only hard dependency: `pip install pyyaml`). Node.js 18+ is needed only if you use the `npx` entry point; the DSH plugin path is pure Python.
- Login-state professional search (ego-search) is off by default and needs a real browser via ego lite (macOS only) or the WebBridge extension; it is not available out of the box and requires extra setup.
MIT · actively maintained (latest release v2.8.2, 2026-08-22)
06Frequently asked questions
How do I install Argo into DeepSeek Harness?
Run `dsh plugin --profile web add "github:taxueseek/argo"` to install the dsh-plugin bundle — it adds 10 `mcp__argo__*` MCP search tools. Optionally add `...#main&path:packages/dsh-plugin` for the wide_research orchestration subpackage, then restart `dsh web`.
Do I need an API key to use it?
No. Argo runs zero-config on free engines plus local `local_*` engines. Adding keys (Tavily, Bocha, etc.) improves quality, and sources without a key are auto-skipped.
What does Argo give a DeepSeek Harness agent that built-in model search doesn't?
It returns compact 'evidence candidate + credibility breakdown' JSON (selection / absorption / freshness / consensus) instead of long prose, with dual-layer caching (~10ms hot queries) and vertical sources for finance, film, sports and academic questions.
Is there a package I should avoid?
Yes — do NOT `npm install argo-search`. That npm copy is an unofficial stale v1.0.1 not maintained by this repo. Install only via `dsh plugin add` or the GitHub `npx` / `install.sh` paths.
What are the 10 MCP tools?
argo_search, argo_local_search, argo_research, argo_evidence, argo_clarify, argo_fetch, argo_crawl, argo_screenshot, argo_pdf and argo_social_search.
07Related DSH workflows
last30days-skill-cn
by jesseovo
AI agent skill that searches China's eight major platforms for the last 30 days of content and compiles a well-sourced research report.
browser4
by platonai
Browser4 — an AI-native browser engine for autonomous agents, intelligent extraction, and large-scale web automation.
dsh-browser
by lum1104
让 DSH 读取/点击/填表/滚动/导航当前 Chrome 标签页（文本优先，保留登录态）
modsearch
by liustack
Web/X 搜索与抓取桥：为纯文本 coding agent 提供结构化 JSON 证据（search/fetch/citations），引擎可选 Antigravity CLI（免 key）/带 key 引擎/X（Grok Build 登录态）。
08Data and sources
dsh plugin --profile web add "github:taxueseek/argo"
主包自带 `dsh.bundle`，`dsh plugin add github:taxueseek/argo` 即得 10 个 MCP 工具；子包再加 `wide_research` 并行研究编排
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-25. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.