Forage

What Forage does

Forage replaces a self-hosted Firecrawl stack, which the README says spins up six containers, with one container running FastAPI, Chromium and extraction libraries. web_search uses Forage's own SERP engines, which render and parse Google, Bing, Yahoo and DuckDuckGo result pages, or optionally SearXNG. web_extract tries static HTTP first and falls back to a browser when a page needs JavaScript or anti-bot handling.

Four browser engines are switchable: scrapling by default, playwright, patchright and chrome-local, which drives the host's desktop Chrome over CDP. Two extract engines, readability and trafilatura, can be set per domain or per request, and a Scrapling solver acts as a last-resort fallback for challenges. The API includes Firecrawl-compatible POST /v1/scrape and POST /v1/search, an in-memory TTL cache, optional Bearer API-key auth and config.yaml settings. It ships with a Hermes plugin implementing WebSearchProvider, and the REST API can also be used standalone.

Key features

• One Docker container with FastAPI, Chromium and extraction libraries
• web_search through its own SERP engines or optional SearXNG
• Hybrid web_extract: static HTTP first, browser fallback for JavaScript and anti-bot pages
• Four browser engines and two extract engines, switchable per config
• Firecrawl-compatible POST /v1/scrape and POST /v1/search
• Ready-made Hermes plugin (WebSearchProvider)

When to use it

• Giving Hermes Agent web search and page extraction without a six-container Firecrawl stack
• Pointing an existing Firecrawl client at a self-hosted backend by changing the base URL
• Extracting content from JavaScript-heavy or anti-bot protected pages

Who it is for: Hermes Agent users who want a lightweight self-hosted web search and extract backend.

How it fits with Hermes Agent

Built specifically for Hermes Agent as its extract and search backend and ships a Hermes plugin; the REST API also works standalone.

Requirements: Docker, since Forage runs as a single container

FAQ