#116116th of 140 in MCP servers for developers
openserp
Self-hosted web search API and MCP server over Google, Bing and other engines
- Stars
- 1.6k
- License
- MIT
- Last commit
- Nov 2026
- Last release
- Nov 2026
- Language
- Go
683 on 25 Apr 20261,598 on 14 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
OpenSERP loads search results for Google, Bing, Yandex, Baidu, DuckDuckGo and Ecosia in Chromium and returns them as one JSON schema with source engine and rank. It exposes a REST API and a CLI, can attach page content as Markdown to the top results, and ships an MCP server plus JS and Python SDKs. No API keys are needed; you supply compute and proxies.
Who it is for: Engineers giving agents or RAG pipelines keyless web search they host themselves
Strengths
- Six engines share one JSON schema; engine and rank survive multi-engine merge
- extract=N attaches Markdown of top results; /extract/batch reads up to 20 URLs
- Single Go binary or Docker image, with proxies, cache, retries and rate limits
- Ready integrations: MCP for Claude Code and Cursor, Open WebUI provider, n8n node
Weaknesses
- Scrapes search pages with a browser, so engine layout changes or blocks can break results
- Raw HTTP mode covers only Google, Yandex, Baidu and Ecosia
- MCP server needs Node.js 18+ and a separately running OpenSERP server
- Reliable Google scraping at volume likely needs your own proxies
What it needs
- no GPU
- Docker + Compose
- Needs Chromium, Node.js 18+ (MCP server), HTTP/SOCKS5 proxies (optional)
- port 7000
What changed
- v0.8.20 ↗ (opens in a new tab)Breaking
“PDF extraction requires mutool from MuPDF on PATH (apt install mupdf-tools on Debian/Ubuntu). The Docker image includes it. PDFs exceeding extract.maxbytes fail explicitly instead of returning partia…”
- v0.8.19 ↗ (opens in a new tab)Breaking
“Process supervisors and scripts now see a non-zero exit code for startup errors that used to log and exit 0.”
- v0.8.17 ↗ (opens in a new tab)Breaking
“Bing web searches make one extra homepage navigation (~0.5s). If the search form fails, the request returns an error instead of falling back to a direct URL. Set bing.searchbox: false or OPENSERPBING…”
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Known problems
Reported, then fixed by a linked change
- bug: DuckDuckGo search returning incorrect results for non-zero offsets ↗ (opens in a new tab)1 reactions · closed 21 Jul 2026Fixed by feat: implement windowOrganicResults function and update DuckDuckGo search logic ↗ (opens in a new tab)
- bug: Remove advertisement entry from DuckDuckGo ↗ (opens in a new tab)0 reactions · closed 26 Oct 2026Fixed by fix(duckduckgo): mark y.js ad links as ads ↗ (opens in a new tab)
- bug: URL extraction ignores HTML <base href> when resolving relative URLs ↗ (opens in a new tab)0 reactions · closed 28 Sept 2026Fixed by fix(extract): honor HTML base href for relative URLs ↗ (opens in a new tab)
Titles are the reporters' own words, taken from the project's GitHub issues and ranked by reactions. We have not tried to reproduce any of them.
Also in MCP servers for developers
See all 140| Rank | Project | Score |
|---|---|---|
| 1 | gitnexusCode knowledge graph exposed to coding agents through MCP tools | 97 out of 100 |
| 2 | codegraphLocal code knowledge graph that coding agents query over MCP | 96 out of 100 |
| 3 | playwright-mcpMCP server that lets LLMs drive a browser through Playwright accessibility snapshots | 93 out of 100 |
| #4 | code-review-graphLocal Tree-sitter code graph that gives coding assistants review context over MCP | 92 out of 100 |
| #5 | fastmcpPython framework for building MCP servers, clients, and interactive apps | 92 out of 100 |