modsearch

Give the web to any model without it: web search, X search, and single-page fetch — free, no signup, no API key.

What problem it solves

Models like DeepSeek-V4-Flash have no web access, or a weak one, so live facts and page fetches are out of reach. ModSearch is a plug-in that greatly strengthens the model’s web search, X search, and single-page fetch. One command routes dsh’s built-in web_search through the modsearch engine chain — no API key, with the native citation cards kept. The default engine is Firecrawl’s keyless tier: 1,000 free credits a month, no account, no card.

It also brings two tools dsh does not ship: x_search for X (Twitter) and read_page for focused single-page reading. With multiple engines configured, failover is automatic and best-first; when a channel fails or runs out of quota, the next one takes over, and meta.attempts records every attempt so a fallback is never silent.

Key features

• Free out of the box: Firecrawl Keyless by default (1,000 free credits / month, no signup, no card); every fallback channel is free too
• Automatic failover: Firecrawl, Antigravity CLI, Tavily, Exa, Grok (X), and local, best-first, with the next engine taking over on failure
• Searches X (Twitter): with Grok Build installed, it queries the corpus web indexes cannot reach
• Single-page reading: read_page fetches and structurally summarizes a URL
• Native citation cards: dsh’s built-in web_search keeps its citation experience on the new engine chain
• Install once, use everywhere: works in Claude Code, Codex, Pi, and OpenCode

Quick start

One command to install (the version is named explicitly so pnpm 11 resolves it correctly):

npx -y @deepseek-ai/dsh plugin —profile web add @liustack/modsearch@5.6.0

It works the moment it lands — no configuration required. To add more free engines, install Antigravity CLI and sign in (curl -fsSL https://antigravity.google/cli/install.sh | bash, then agy), or add a personal quota with one line: modsearch config set tavily.apiKey <key>. Keys live in ~/.modsearch/config.json (mode 0600, masked when shown) and can also come from the TAVILY_API_KEY, EXA_API_KEY, and FIRECRAWL_API_KEY environment variables.

Who it is for

• Researchers and developers running no-web (or weak-web) models on DSH who need reliable retrieval
• Users who want X search and single-page reading without paying for search services
• Users who want redundant engines so retrieval keeps working when one channel fails or exhausts quota