OpenSearch¶
@kagan-sh/opensearch is an evidence-backed search tool for AI coding agents.
It searches the live web and public code in parallel, then returns structured JSON your agent can act on. Works as an OpenCode plugin and as a Claude Code MCP server.
Get started¶
| Runtime | Install guide |
|---|---|
| Claude Code (MCP server) | Claude Code guide |
| OpenCode (plugin) | Quickstart |
Both paths need a SearXNG instance for web search. Code search works out of the box.
Reference¶
| Topic | Page |
|---|---|
| Environment variables and config shapes | Configuration |
| JSON response shape | Result contract |
| Built-in sources and extension status | Sources |
| LLM skill prompt | SKILL.md |
| Contributing | Contributing |
| Feature requests | Feature requests |
What it is good at¶
- broad research before implementation
- comparing signals across internal and external evidence
- collecting citations for a structured answer
What it is not¶
- a general-purpose search engine
- a dynamic plugin ecosystem for arbitrary datasources
- a substitute for domain-specific ranking or private enterprise search
Today, new datasources still require changes in core. That is intentional for now: one small tool, one obvious path, minimal moving parts.