Skip to content
BSWEN
Blog

Best MCP Servers for OpenCode CLI: Tavily, Searxng, and DCP Plugins Compared

Cowrie
Cowrie
Dev @ Bswen

I wanted OpenCode CLI to match Claude Code’s research performance, but I quickly realized it couldn’t access anything beyond its training data. Every time I asked about recent library updates or current documentation, I got outdated or hallucinated answers.

The problem? OpenCode CLI doesn’t have built-in web access. It’s stuck with whatever knowledge was baked into its training cutoff.

The Research Gap in OpenCode CLI

When I first started using OpenCode, I assumed it could search the web like Claude Code. I was wrong.

I asked it about the latest React 19 features, and it gave me information about React 18. I asked about a bug fix released last week, and it had no idea what I was talking about. The tool was essentially operating in the past.

This is where MCP (Model Context Protocol) servers come in. They’re external tools that give OpenCode access to real-time data sources - web search, documentation databases, and custom APIs.

Tavily MCP: The Managed Solution

After digging through Reddit discussions, the Tavily MCP server kept coming up as the top recommendation. One comment with 12 upvotes said: “Just add the Tavily MCP for web search and scraping, otherwise it can only access the research it was trained on.”

I set it up and immediately saw the difference. Tavily is a search API specifically designed for AI applications. It doesn’t just dump raw HTML - it returns structured, LLM-optimized results.

Configuration

mcp_settings.json
{
  "mcpServers": {
    "tavily": {
      "command": "npx",
      "args": ["-y", "@tavily/mcp-server"],
      "env": {
        "TAVILY_API_KEY": "tvly-your-key-here"
      }
    }
  }
}

You’ll need to get an API key from tavily.com. They have a free tier with 1,000 searches per month, which is enough for personal use.

What Tavily Does Well

  • Returns clean, structured search results optimized for LLMs
  • Includes web scraping capabilities for detailed content extraction
  • Managed service - no infrastructure to maintain
  • Good documentation and active development

The Trade-offs

  • Requires an API key and external service dependency
  • Free tier limits may not be enough for heavy users
  • Your queries go through a third-party service

Searxng MCP: The Privacy-First Alternative

If you’re privacy-conscious or want full control over your search infrastructure, Searxng MCP is the answer. As one Redditor noted: “I’m doing it with a selfhosted Searxng MCP server for example.”

Searxng is a metasearch engine that aggregates results from multiple search engines while respecting your privacy. It doesn’t track you, doesn’t profile you, and you host it yourself.

Setting Up Searxng

docker-compose.yml
version: "3"
services:
  searxng:
    image: searxng/searxng:latest
    ports:
      - "8888:8080"
    environment:
      - SEARXNG_BASE_URL=http://localhost:8888/
      - SEARXNG_SECRET=your-secret-key-here
    volumes:
      - ./searxng:/etc/searxng

Start it with:

terminal
docker-compose up -d

Then configure OpenCode to use it:

mcp_settings.json
{
  "mcpServers": {
    "searxng": {
      "command": "npx",
      "args": ["-y", "@searxng/mcp-server"],
      "env": {
        "SEARXNG_URL": "http://localhost:8888"
      }
    }
  }
}

Why Choose Searxng

  • Complete privacy - your data never leaves your server
  • No API limits or quotas
  • Aggregates results from multiple search engines
  • You control the infrastructure

The Trade-offs

  • Requires setup and maintenance
  • You’re responsible for updates and security
  • May need more resources to run

DCP Plugin: The Unique Research Tool

The DCP plugin caught my attention with this comment: “DCP is the very best plugin right now, there is nothing compared to it on other AI Harness.”

While it only had 1 upvote in the discussion, the description suggests it offers capabilities beyond simple web search. DCP appears to be a comprehensive research enhancement tool that’s unique to the OpenCode ecosystem.

I haven’t tested this one personally yet, but based on community feedback, it seems to provide deeper research capabilities than basic search APIs. If you need more than just web search - perhaps document analysis, cross-referencing, or specialized data sources - DCP might be worth exploring.

Where MCP Settings Live

One thing that confused me initially was where to configure these MCP servers. The location depends on your setup:

terminal
# Global configuration (applies to all projects)
~/.config/opencode/mcp_settings.json


# Project-specific configuration
./.opencode/mcp_settings.json

Project-specific settings override global settings, so you can have different MCP configurations for different projects.

Context7 MCP: For Official Documentation

While researching MCP servers, I also came across Context7 MCP. This one is specifically designed for fetching official documentation, which is useful when you need accurate, up-to-date library references.

If you’re working with rapidly evolving frameworks or libraries, Context7 can help OpenCode access the latest official docs instead of relying on potentially outdated training data.

Common Mistakes I Made

  1. Assuming OpenCode had built-in web access - It doesn’t. You need MCP servers for any external data access.

  2. Not testing MCP integration - I spent hours wondering why my searches weren’t working, only to realize I had a typo in my configuration file.

  3. Ignoring API costs - Free tiers are great for testing, but if you’re using OpenCode daily, you might hit limits quickly.

  4. Skipping the documentation - Each MCP server has its own configuration quirks. Reading the docs saves hours of debugging.

Choosing the Right MCP Server

Here’s a quick decision guide:

  • Tavily MCP if you want something that just works and don’t mind using a third-party service
  • Searxng MCP if you prioritize privacy and have the resources to self-host
  • DCP Plugin if you need advanced research capabilities beyond basic web search
  • Context7 MCP if you primarily need access to official documentation

You can also combine multiple MCP servers. For example, I use Tavily for general web search and Context7 for library documentation.

The Bottom Line

OpenCode CLI is powerful, but it needs external tools to reach its full potential. Without MCP servers, you’re limited to outdated training data. With the right MCP configuration, OpenCode can access real-time information and match Claude Code’s research performance.

The Reddit community has spoken: Tavily MCP is the most popular choice, Searxng offers privacy-focused alternatives, and DCP provides unique research capabilities. Pick based on your needs, set it up once, and enjoy the upgraded research experience.

Final Words + More Resources

My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me

Here are also the most important links from this article along with some further resources that will help you in this scope:

Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!

Comments

Tags: