Firecrawl
NewThe free alternative toContext.dev
Firecrawl is the open source alternative to Context.dev: it turns any website into clean Markdown for Claude and AI agents. Give it a page, it returns the text without menus or ads. It also crawls a whole site, lists its URLs, searches the web and, with an AI model (OpenAI or a local model through Ollama), pulls structured data out of it. At home, it runs with Docker Compose under the AGPL-3.0 license: free, with no credit meter, and connected to Claude through its MCP server. Their cloud gives 1,000 credits a month, then about €17 a month (19 dollars). The limit: their anti robot engine, proxy rotation, screenshots and actions (clicking, filling) stay reserved for the cloud. On a protected site, you need your own proxies, which are paid.
Versus Context.dev: the match
Price
Firecrawl
€0 on your machine
Context.dev
€18/month (Developer)
Pages a month
Firecrawl
Unlimited
Context.dev
10,000 (1,000 on the free plan)
Protected sites (anti robot, proxies)
Firecrawl
Your own proxies, paid
Context.dev
Included, even on the free plan
| VS | Firecrawl | Context.dev ↗ |
|---|---|---|
| Price | €0 on your machine | €18/month (Developer) |
| Pages a month | Unlimited | 10,000 (1,000 on the free plan) |
| Protected sites (anti robot, proxies) | Your own proxies, paid | Included, even on the free plan |
| Screenshots | Reserved for Firecrawl's cloud | Included |
| Whole site and web search | Included, no meter | From 1 credit per page or search |
| AI extraction (JSON) | OpenAI or a free local model (Ollama) | 5 credits per page |
| What you need | Docker and a machine that can run several services | An API key |
How to install it (even from zero)
Several Docker services (the API, a browser, a database and queues), on your computer or your server. The prompt installs the free version from the official repo, checks a first page and connects Firecrawl to Claude.
- 1
Paste this prompt into Claude (Code or Desktop)
It installs Firecrawl from the official repo, checks that a page comes out as Markdown and connects it to Claude through the MCP server:
Installe Firecrawl sur ma machine : c'est l'outil open source qui transforme des sites web en Markdown propre pour les agents IA, l'alternative à Context.dev (dépôt GitHub firecrawl/firecrawl, documentation docs.firecrawl.dev). Vérifie d'abord que Git, Docker et Docker Compose sont installés, que Docker a assez de mémoire pour faire tourner plusieurs services, et que le port 3002 est libre. Suis ensuite le guide d'auto-hébergement de la documentation (docs.firecrawl.dev/contributing/self-host) : clone le dépôt, place-toi sur la version qu'il a vérifiée, crée le fichier .env minimal qu'il donne en remplaçant le mot de passe PostgreSQL par une valeur longue et aléatoire, puis construis et lance l'ensemble avec Docker Compose. Vérifie que l'API répond sur http://localhost:3002 en récupérant une page publique en Markdown. Branche ensuite Firecrawl à Claude avec le serveur MCP officiel (dépôt GitHub firecrawl/firecrawl-mcp-server), en réglant FIRECRAWL_API_URL sur l'adresse de mon API. Si je veux extraire des données structurées, propose-moi Ollama pour un modèle local gratuit, ou ma clé OpenAI. L'API n'a pas de mot de passe par défaut : si c'est sur un serveur, ne l'ouvre pas sur Internet sans authentification ni HTTPS. Explique-moi aussi quoi faire si un site bloque Firecrawl (proxies), et comment le mettre à jour. Explique chaque étape simplement, je débute. Review the GitHub repository and check whether it is safe to install given my machine and my infrastructure. Ask me as many questions as possible, using widgets, to clarify my request and be as precise as you can.
- 2
Start with a public site
Test on a site with no anti robot protection: that is where the free version does everything. A protected site needs proxies, or their cloud.
- 3
Keep the API to yourself
By default, the API answers without a password. Keep it on your machine, or put it behind your domain name over HTTPS, with protected access, before opening it to the Internet.
Paste this prompt into Claude (Code or Desktop)What you actually do with it
Give Claude the web
With the MCP server, Claude reads any page as Markdown, searches the web and crawls a whole site, without subscribing to an API.
Turn a site into a knowledge base
Firecrawl crawls every page of a site and returns it as Markdown: enough to feed a chatbot, a second brain or a Claude project.
Pull structured data
Describe the fields you want (price, title, date) and the AI model pulls them from each page as JSON. With Ollama, it is free.
Project health, at a glance
Before installing an open-source tool, one question matters: is the project alive? These scores are computed from public GitHub data. They are not made up.
93/100
ExcellentVery active, well-maintained project: go ahead with confidence.
- Activity
- 100
Is the code moving? (commits in the last 30 days)
- Maintenance
- 100
How fresh is the latest update? Fresher is better.
- Community
- 100
Stars and contributors: people around means help and fixes.
- Maturity
- 80
Project age. Old enough to be solid, still alive.
- Momentum
- 85
Current pace versus its usual pace: speeding up or dozing off?
The repo in numbers
- GitHub stars
- 196.6k
- Forks
- 10.4k
- Contributors
- 167
- Commits
- 6.7k
- Project age
- 2.5 yrs
- Last commit
- today
- Size
- 183.3 MB
- License
- AGPL-3.0
Built with
- TypeScript 72.8%
- Python 11.9%
- Rust 3.4%
- Java 2.9%
- PHP 2.3%
Scores computed automatically from the GitHub API (last-30-days activity, update freshness, community size, project age). Refreshed daily.
Loucash’s take
7/10
Firecrawl does the same job as Context.dev: a page, a whole site or a web search, turned into clean Markdown for Claude and your agents. At home, it is free with no credit meter. What I like: the project is closely followed and updated almost every day, the code is under a real open source license (AGPL-3.0), and the official MCP server connects it to Claude with no subscription. You can use it for yourself or your company without publishing anything: only someone who modifies it and offers it online to others has to share their code. Why 7 and not more: the free version lacks what their cloud charges for. The anti robot engine, proxy rotation, screenshots and actions stay with them, while Context.dev includes proxies even on its free plan. And the install runs several Docker services, with a password free API that you have to protect yourself.
It’s for you if: You want to give Claude or your agents the web (reading pages, crawling sites, searching) without paying per credit, and Docker does not scare you. If you need to collect heavily protected sites at scale, or take screenshots, Context.dev or Firecrawl's cloud will still be needed.
Also worth a look in the library
OpenSEO
Data & analytics€134/moThe open source SEO tool that does the job of Semrush and Ahrefs: keyword research, backlinks, rank tracking, site audit, visibility in AI answers. The software is free and self-hosted; the SEO data is pay as you go, a few cents per request, no subscription.
View details →
Casdoor
Data & analytics€31/moThe open source sign in server that replaces Auth0: a login page in your colours, sign in with Google and about a hundred others, two factor authentication, enterprise SSO. Your user accounts stay on your server, with no bill that grows with them.
View details →
PocketBase
Data & analytics€22/moA complete backend in a single file: SQLite database, user accounts, files, realtime and an admin dashboard. You run it on your server and get the equivalent of Supabase or Firebase, with no subscription.
View details →
One email a week that saves you money
Every Tuesday: the open source alternatives that just entered the library, what they replace and what they save you. One email, nothing more.
Free. One email on Tuesdays. Unsubscribe in one click. Your data
Join the Tuesday readers
