mcp-searchable
mcp-searchable is a local-first Model Context Protocol (MCP) server that helps AI agents perform practical web research:
- Search with multiple providers (
duckduckgo,google,bing,brave) - Fetch pages and convert them into readable Markdown
- Stash useful content into local SQLite storage
- Grep stashed content with FTS5
- Ask questions over local context using Ollama
It is built for reproducible, tool-based research flows where the agent can both gather and reuse context.
What does it do?
- Standardizes web search output across providers into
{ title, url, snippet } - Extracts readable content from raw web pages
- Persists page content locally so agents can reuse it later
- Enables local semantic workflows without depending on hosted vector services
Why mcp-searchable?
Most agent workflows lose context between searches and page reads. mcp-searchable gives agents a persistent local knowledge loop:
- discover sources (
web_search) - extract content (
web_fetch) - store useful material (
web_stash) - retrieve relevant context (
web_grep) - synthesize answers (
web_ask)