SourceWeave Web Search

Search-first MCP server and CLI for web research.

Overview • Getting started • Managed local runtime • MCP client setup • CLI • Container deployments • OpenWebUI • Runtime configuration • Development

Overview

SourceWeave Web Search gives MCP clients a compact three-tool contract for web research:

• search_web(query, domains?, urls?, effort?) discovers sources and returns compact results with stable page_id handles.
• read_pages(page_ids, focus?) reads stored pages by page_id.
• read_urls(urls, focus?) reads direct URLs without searching first.

It combines:

Getting started

Requirements

• Python 3.12+
• Docker with Compose support for the default managed local runtime
• Explicit SOURCEWEAVE_SEARCH_* endpoints only if you want hosted or self-managed services

Managed local runtime

Run the server from the published package:

uvx --from sourceweave-web-search sourceweave-search-mcp

Or start the MCP server over HTTP:

uvx --from sourceweave-web-search sourceweave-search-mcp \
  --transport streamable-http \
  --host 127.0.0.1 \
  --port 8000

When no endpoint env vars are set, sourceweave-search-mcp:

Managed state lives under ~/.sourceweave-local/managed-runtime. Multiple MCP processes on the same machine share one managed stack per state directory.

Explicit endpoint mode

If you already run SearXNG, Crawl4AI, and Redis or Valkey yourself, or want to point at hosted services, set explicit endpoints and the MCP entrypoint will bypass managed Docker startup:

SOURCEWEAVE_SEARCH_SEARXNG_BASE_URL="http://127.0.0.1:19080/search?format=json&q=<query>" \
SOURCEWEAVE_SEARCH_CRAWL4AI_BASE_URL="http://127.0.0.1:19235" \
SOURCEWEAVE_SEARCH_CACHE_REDIS_URL="redis://127.0.0.1:16379/2" \
uvx --from sourceweave-web-search sourceweave-search-mcp

Direct CLI

sourceweave-search runs the tool directly. Use it when the supporting services are already available or when you provide explicit endpoints. It does not start Docker.

sourceweave-search --query "python programming" --read-first-pages 2
sourceweave-search --read-url "https://packaging.python.org/en/latest/"

MCP client setup

OpenCode

Example opencode.json / opencode.jsonc / ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "sourceweave": {
      "type": "local",
      "command": [
        "uvx",
        "--from",
        "sourceweave-web-search",
        "sourceweave-search-mcp"
      ],
      "enabled": true,
      "timeout": 300000
    }
  }
}

For a shared HTTP endpoint instead:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "sourceweave": {
      "type": "remote",
      "url": "http://127.0.0.1:18000/mcp",
      "enabled": true,
      "timeout": 300000
    }
  }
}

VS Code Copilot

Example .vscode/mcp.json:

{
  "servers": {
    "sourceweave": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "--from",
        "sourceweave-web-search",
        "sourceweave-search-mcp"
      ]
    }
  }
}

For a shared HTTP endpoint instead:

{
  "servers": {
    "sourceweave": {
      "type": "http",
      "url": "http://127.0.0.1:18000/mcp"
    }
  }
}

Claude Code

Example .mcp.json:

{
  "mcpServers": {
    "sourceweave": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "--from",
        "sourceweave-web-search",
        "sourceweave-search-mcp"
      ]
    }
  }
}

For a project-scoped shared config, place the same block in .mcp.json at the repo root.

CLI

The direct CLI is useful once the supporting services are already reachable. It gives you the same search-first workflow without the MCP wrapper.

sourceweave-search --query "react useEffect cleanup example" --read-first-page
sourceweave-search --query "HTTP overview" --domain developer.mozilla.org --read-first-page
sourceweave-search --read-url "https://packaging.python.org/en/latest/"

Container deployments

The managed local runtime is for host-side uvx or uv run launches. Containerized deployments still use explicit endpoint wiring.

• Image: ghcr.io/mrnaqa/sourceweave-web-search-mcp
• Repo-local compose entrypoint: docker compose up -d --build mcp

Example container run:

docker run --rm -p 8000:8000 \
  -e SOURCEWEAVE_SEARCH_SEARXNG_BASE_URL="http://host.docker.internal:19080/search?format=json&q=<query>" \
  -e SOURCEWEAVE_SEARCH_CRAWL4AI_BASE_URL="http://host.docker.internal:19235" \
  -e SOURCEWEAVE_SEARCH_CACHE_REDIS_URL="redis://host.docker.internal:16379/2" \
  ghcr.io/mrnaqa/sourceweave-web-search-mcp:latest

OpenWebUI

This repo also ships a generated standalone OpenWebUI tool file at artifacts/sourceweave_web_search.py.

From a repo checkout, verify it is in sync with the canonical implementation:

uv run sourceweave-build-openwebui --check

Paste that artifact into OpenWebUI when you want the standalone tool-file deployment path. The generated file rewrites the default endpoints to the repo-local compose service names so it matches the container deployment path out of the box.

Runtime configuration

Optional environment variables:

If the endpoint variables are unset, sourceweave-search-mcp defaults to managed local runtime.

• Canonical host endpoints remain the preferred defaults and the external-reuse probe targets.
• A SourceWeave-managed stack may use different free host ports when the canonical defaults are already occupied.
• Multiple MCP processes on the same machine share one managed stack per local runtime state directory.

Default endpoint values:

• SearXNG: http://127.0.0.1:19080/search?format=json&q=<query>
• Crawl4AI: http://127.0.0.1:19235
• Redis: redis://127.0.0.1:16379/2

Default preferred host ports for managed startup:

• SearXNG: 19080
• Crawl4AI: 19235
• Redis: 16379
• MCP: 8000 when run directly with uvx; 18000 at /mcp when using the repo's mcp compose service

Development

git clone https://github.com/MRNAQA/sourceweave-web-search.git
cd sourceweave-web-search
uv sync --locked --group dev
uv run sourceweave-search-mcp

Useful checks:

uv run sourceweave-build-openwebui --check
uv run sourceweave-search-mcp --help
uv run pytest tests/test_config.py tests/test_packaging.py tests/test_tool.py tests/test_managed_runtime.py -m "not integration"