CommunityActive· 6d★ 4.3by Ihor Sokoliuk

SearXNG MCP

Connects to any SearXNG instance for web search and URL content fetching with pagination and time filtering.

View repo

Safety & Trust

Delv Safety Grade: C

Score 61/100 · assessed 2026-04-28

Maintainer45

Permissions75

Supply chain65

Transparency70

Incidents100

SearXNG MCP is a community server by solo developer Ihor Sokoliuk that connects Claude to self-hosted SearXNG metasearch instances. The privacy angle is genuine: you control the search backend and avoid tracking from commercial engines. However, this is a single-maintainer project with limited track record. The package is distributed via npm with npx execution, which is standard but requires trusting the maintainer's registry account. Permissions are reasonably scoped to network outbound (hitting your SearXNG URL) and reading one environment variable. The main risk is maintainer bus factor and the fact that you're routing all Claude's search queries through code from an individual developer with no organisational backing. Transparency is adequate with open source on GitHub, but documentation and community engagement appear thin. No known incidents, but limited scrutiny given the small user base.

Lethal Trifecta (prompt-injection exposure)

TWO OF THREE

Private dataNo

Reads secrets, credentials, private files

Untrusted inputYes

Ingests web pages, PRs, issues, emails

External commsYes

Can send data outbound

Self-hosted meta-search. Same axes; you choose your upstreams.

Green flags

Privacy-focused design using self-hosted search infrastructure
Scoped to read-only search operations, no filesystem or shell access
Open source with visible code on GitHub
Standard npm distribution via npx

Red flags

Solo maintainer with no organisational backing or track record
Requires SEARXNG_URL env var which could point to malicious endpoints
Limited community review or adoption signals
No versioning discipline visible in repo metadata

Permissions requested

Outbound networkRead env

Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Install

npx -y mcp-searxng

Env vars needed: SEARXNG_URL

Review

SearXNG MCP connects Claude to any SearXNG instance, which is a self-hosted metasearch engine that aggregates results from dozens of sources without tracking you. You point it at your own SearXNG server (or a public one, though that defeats the privacy angle), and Claude can search the web or fetch URL content on your behalf. I'd reach for this if I'm already running SearXNG and want Claude to have search capabilities without phoning home to Google or Bing. The privacy story is the main draw here. You control the search instance, you control the logs, and you can configure which upstream engines SearXNG queries. The MCP supports pagination, so Claude can walk through multiple pages of results if the first batch isn't enough. Time filtering is in there too, so you can scope searches to recent content. The URL fetching is handy for pulling down page content without leaving Claude. It's not a full-featured scraper, but for grabbing article text or checking what's on a page, it does the job. I've used it to fact-check claims in a conversation by searching my SearXNG instance and then fetching the top result in one go. Quirks: you need a SearXNG instance already running. If you don't have one, this MCP is pointless until you spin one up, which is a separate project. The setup assumes you know your instance's URL and have it accessible from wherever Claude is running. No built-in rate limiting or error handling beyond what SearXNG itself provides, so if your instance is slow or down, Claude will just wait or fail. Skip this if you're happy with Claude's built-in search or don't care about privacy. Also skip it if you're not technical enough to run a SearXNG instance. This is for people who already have the infrastructure or are willing to build it. For everyone else, it's overkill.

Verdict

Install this if you run your own SearXNG instance and want Claude to search without leaking queries to commercial engines. Skip it if you don't have SearXNG already or if privacy isn't a priority. It's a single-purpose tool that does one thing well, but only if you've done the groundwork.

Good at

Full control over search privacy since you host the SearXNG instance yourself.
Pagination support means Claude can fetch multiple pages of results without manual intervention.
Time filtering lets you scope searches to recent content, which is useful for news or trending topics.
URL fetching is built in, so Claude can grab page content in the same workflow.
Works with any SearXNG instance, public or private, as long as you have the URL.

Watch out

Requires a running SearXNG instance, which is a separate setup and maintenance burden.
No built-in rate limiting, so you're at the mercy of your SearXNG instance's performance.
Config snippet isn't provided in the repo, so you'll need to piece it together from the README.
Only tested with Claude Desktop and a few other hosts, so broader compatibility is unclear.

Use cases

self-hosted search
private web research
aggregated metasearch
uncensored queries

Getting started

1. Spin up a SearXNG instance if you don't have one already. The official Docker image is the easiest route. 2. Run `npx -y mcp-searxng` and add the server block to your Claude Desktop config, setting `SEARXNG_URL` to your instance's base URL (e.g., `http://localhost:8080`). 3. Restart Claude Desktop and ask it to search for something specific, like "recent Node.js security advisories". It should query your SearXNG instance and return results. 4. Check the SearXNG logs to confirm queries are hitting your instance, not a third-party service. 5. Watch out: if your SearXNG instance is behind auth or uses non-standard ports, you'll need to adjust the URL and possibly add credentials.

Works with

Claude DesktopClaude CodeCursor