How to self-host a Tavily alternative

In short: clone Siftdog, run docker compose up, and you have the Tavily API on your own machine.

1. Start it

git clone https://github.com/khsarvar/siftdog
cd siftdog
docker compose up

This starts the API on port 8000 and a SearXNG instance for search results. OpenAPI docs are at http://localhost:8000/docs.

2. Try a search

curl -s localhost:8000/search -H "Content-Type: application/json" \
  -d '{"query": "latest python release", "search_depth": "advanced"}'

basic returns search snippets. advanced reads each page and returns its most relevant passages.

3. Point your Tavily code at it

from tavily import TavilyClient

client = TavilyClient(api_key="your-siftdog-key", api_base_url="http://localhost:8000")
client.search("latest python release")

The same works for extract, crawl and map. See using Siftdog with LangChain and the Tavily SDK.

4. Before exposing it

• Set API_KEYS in .env (comma-separated). Clients then send Authorization: Bearer <key>.
• Keep ALLOW_PRIVATE_NETWORKS=false. /extract and /crawl fetch URLs your callers choose, so private and internal addresses are blocked on every redirect and at connect time.
• For the MCP endpoint behind a domain, add the hostname to MCP_ALLOWED_HOSTS.

What to expect

In our benchmark, both found the answer to all 25 factual questions in advanced mode, and Siftdog returned official documentation more often.

Would rather not run servers? Join the hosted waitlist.