AI Agent Hub
Back to plugins
🧰

dsh-exa-search

Web Tools Updated 2026.08.21

Run the following command in DeepSeek Harness:

dsh plugin install Kev-ZJY/dsh-exa-search

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install Kev-ZJY/dsh-exa-search in the DeepSeek Harness terminal; the source is at https://github.com/Kev-ZJY/dsh-exa-search, then set EXA_API_KEY to activate the exa_search tool.

About this plugin

DeepSeek Harness ships with a built-in web_search that works well for quick lookups, but when a single research pass demands titles, links, authors, dates, and abstracts all at once, the results can feel thin. dsh-exa-search registers a dedicated exa_search tool backed by the Exa /search API, exposing six search types—auto, fast, instant, deep-lite, deep, and deep-reasoning—so one call can return structured, multi-source results. It complements rather than replaces web_search: quick link check goes to web_search, deep retrieval goes to exa_search.

The plugin puts guardrails around every request. Query length, result count, response body size, and model-visible output are all capped. Request cancellation and timeouts are supported. API keys never appear in request logs and are flagged as sensitive in configuration. A fixed set of stable error codes (EXA_TIMEOUT, EXA_RESPONSE_TOO_LARGE, EXA_INVALID_ARGUMENT, and more) makes upstream error handling and log analysis predictable. The tool is model-agnostic with no GPU runtime dependency, and every setting can be overridden via environment variables, plugin configuration, or bundle patches.

If your day-to-day workflow leans on DSH for technical research, literature discovery, or cross-source verification and you want a search tool that can shift strategy based on task depth without locking you into a single vendor, dsh-exa-search is a drop-in option. Set an Exa API key, restart the profile, and exa_search will be waiting in the available tool list.

Use Cases

  • Tech research or literature discovery where a single call must return multi-source titles, links, and abstracts
  • Vendor models that lack Anthropic protocol support but still need DSH tool-calling for search
  • Built-in web_search is too shallow and you need Exa deep or deep-reasoning for richer retrieval

Best For

  • Researchers who rely on DSH for daily deep reading and technical research
  • Developers using non-Anthropic vendor models that still need tool-calling search
  • Teams that want search depth to adapt to task type without vendor lock-in