Skip to content

Supercharge your AI agents with data from the web and beyond. Building the library for superintelligence. 🔥

add claude file

Gergő Móricz3f0873c78882
diff --git a/CLAUDE.md b/CLAUDE.md
new file mode 100644
index 000000000..8b3a89a96
--- /dev/null
+++ b/CLAUDE.md
@@ -0,0 +1,19 @@
+Firecrawl is a web scraper API. The directory you have access to is a monorepo:
+ - `apps/api` has the actual API and worker code
+ - `apps/js-sdk`, `apps/python-sdk`, and `apps/rust-sdk` are various SDKs
+
+When making changes to the API, here are the general steps you should take:
+1. Write some end-to-end tests that assert your win conditions, if they don't already exist
+ - 1 happy path (more is encouraged if there are multiple happy paths with significantly different code paths taken)
+ - 1+ failure path(s)
+ - Generally, E2E (called `snips` in the API) is always preferred over unit testing.
+ - In the API, always use `scrapeTimeout` from `./lib` to set the timeout you use for scrapes.
+ - These tests will be ran on a variety of configurations. You should gate tests in the following manner:
+ - If it requires fire-engine: `!process.env.TEST_SUITE_SELF_HOSTED`
+ - If it requires AI: `!process.env.TEST_SUITE_SELF_HOSTED || process.env.OPENAI_API_KEY || process.env.OLLAMA_BASE_URL`
+2. Write code to achieve your win conditions
+3. Run your tests using `pnpm harness jest ...`
+ - `pnpm harness` is a command that gets the API server and workers up for you to run the tests. Don't try to `pnpm start` manually.
+ - The full test suite takes a long time to run, so you should try to only execute the relevant tests locally, and let CI run the full test suite.
+4. Push to a branch, open a PR, and let CI run to verify your win condition.
+Keep these steps in mind while building your TODO list.
\ No newline at end of file