Use your agent to build any of their products.
RAE FAANG catalogs products from Meta, Amazon, Apple, Netflix and Google. Choose a product and use your coding agent to build it in your project. Each entry links product context, research, candidate open-source projects and product-specific build instructions to help you get started. These are starting points, not limits on what you can build. RAE FAANG is MIT-licensed; your implementation and its dependencies have their own licenses.
npx rae-faang products list
Live site · Choose a product to build · Contributing · Issues
Important
RAE FAANG is a product catalog for agent-led building, not a source-code scraper. It brings together public product information, research and candidate open-source projects so your coding agent has useful starting points for building in your project. The research does not limit what you can build. See Responsible reverse engineering.
- Choose any product in the live catalog or the
catalog/skills/index. - Give its Markdown skill to your coding agent from the project where you want to work.
- Ask the agent to check the research, assess candidate projects, then plan, build and test a scoped implementation. Review its choices and the test results.
Build a search product for my project. Start from the google-search skill
in the RAE FAANG catalog. Assess open-source candidates such as SearXNG
first, then scope, build, and verify what fits my users.
No clone is needed for the catalog itself — the published npm package bundles the catalog, research dossiers, and open-source landscape.
Use a skill without installing anything
npx rae-faang products list # every catalogued product
npx rae-faang research coverage # how researched is the catalog?
npx rae-faang gaps all # what is missing?
The npm package exposes both rae and rae-faang executable names. For a
one-off command, invoke the published package as npx rae-faang ...;
after npm install -g rae-faang, use rae ....
Generate a product skill from npm
npx rae-faang google-search --out ./google-search.md
This product-ID shorthand writes one portable Markdown skill to the path you
choose (or ./<product-id>.md by default). Give that file to your coding agent
in the project where you want to build. For repeated use, install globally with
npm install -g rae-faang, then run rae google-search --out ./google-search.md.
The longer skills generate <product-id> form is also available for batch
runs and JSON reports. skills validate checks the bundled canonical skills;
run npm run validate from a source checkout to check repository output drift.
If you specifically want a directory containing both skill.md and mcp.json,
use npx rae-faang build agent-kit gmail --out ./gmail-agent.
Work on the catalog itself (contributors)
git clone https://github.com/EnzoVezzaro/rea-faang.git rae-faang
cd rae-faang
npm install
npm test # CLI + schema test suite
npm run validate # research + alternatives + skills + site build check
npm run site:build # regenerate the landing/catalog site
ace status # the agent environment
Tip
Keep npm run validate green before every commit. It is the gate the CI runs.
A product build can follow these five steps. The skill provides a starting point; you choose the scope and review the work.
| Stage | What happens | Your control |
|---|---|---|
| 1. Investigate | Review public product information, official documentation and published protocols | Confirm which user workflows matter |
| 2. Plan | Turn findings into features, an architecture and checks for the chosen scope | Approve or adjust the plan |
| 3. Build | Implement the agreed features using suitable technology and independently written code | Choose the stack and review changes |
| 4. Test | Check the implementation against its requirements, including error cases and security | Run or review the tests |
| 5. Ship or continue | Update the documentation and decide whether to release, keep working or stop | Decide what happens next |
Alternative-coverage labels (available, partial, no verified alternative) describe the
catalog's research into existing open-source projects. They do not tell you whether your own
implementation is feasible.
| Company | Product families catalogued |
|---|---|
| Meta | Social networking, messaging, communities, media sharing, collaboration, VR software |
| Amazon | E-commerce, marketplaces, cloud infrastructure, developer services, streaming, digital assistants |
| Apple | Desktop and mobile experiences, productivity, creative applications, media services, device ecosystems |
| Netflix | Video streaming, content discovery, recommendation systems, media delivery, creator tooling |
| Search, maps, email, video, office productivity, cloud infrastructure, developer tools, AI services |
77 products · 40 with a mature candidate listed · 29 with partial coverage · 8 with
no verified alternative. These labels summarize catalog research, not product parity. See
catalog/products.jsonl and the live catalog for current counts.
catalog/— 77 product records with JSON Schemas, research manifests, and the REA web-surface reports captured withREA .catalog/skills/— one Markdown build skill per product, from the skills index catalog down to the per-product file.research/companies/— one evidence dossier per product, with citations and unknowns.src/cli.js— the deterministic CLI that validates research, tracks coverage, and reports gaps. No SDK dependencies beyondminimistandjs-yaml.tools/— catalog and site generators (skills, alternatives map, landing/catalog build).site/— the generated landing page and catalog, deployed to GitHub Pages.- ACE (
ace-proagents) — an optional agent-orchestration environment per product workspace.
The repository is designed to be forked, extended, and shipped in increments.
Where to start as a contributor
- Read
CONTRIBUTING.md— setup, the contribution loop, and review gates. - Read
AGENTS.mdif you work with AI agents — roles, task leases, handoffs. - Pick a gap:
npx rae-faang gaps allprints the largest shortfalls by coverage type. - Improve one dossier, one alternative entry, or one test — small diffs get reviewed faster.
Project conventions that make contributions land
- Cite public sources for material research claims in
evidence[]; record uncertainty where a source is missing or unclear. - Clean room: no decompilation, proprietary source, protected assets, or access-control bypass.
- The CLI is the arbiter:
npm run validateandnpm testmust pass to merge. - Verification is never weakened to pass a check — fix the cause instead.
- Docs in
docs/are canonical;.docs/is historical only.
Versioning and releases
Releases follow semver and are recorded in
CHANGELOG.md. Catalog improvements bump the patch or minor
version; the maintainer publishes to npm after CI is green. Each release is
tagged in this repository.
Do I need the ACE agent environment? #
No. ACE (ace-proagents) is optional orchestration for people who want a coordinated agent
team. One coding agent reading one product skill is enough to start a build.
Which coding agents can use a product skill? #
Any agent that can read a Markdown file. The skill is plain Markdown with product context, candidate projects, and build/verification instructions — no MCP server or special runtime is required.
Do I have to build all 77 products? #
No. Pick one product, scope it to your needs, and build that. The catalog's research gives your agent where to start; your project defines how far it goes.
What does the coverage label mean for my build? #
It summarizes research into existing open-source projects for a product. It does not decide
what your agent can build: a no verified alternative product is a research gap, not proof
that building it is impossible — the skill says so explicitly.
How do I report a problem or add a product? #
Open an issue. For missing
alternatives, use npx rae-faang gaps all output to point at the exact coverage
gap. For new products, see the catalog section of CONTRIBUTING.md.
What should I do if a skill makes a claim I can't verify? #
That's a finding, not a nuisance — open the dossier in research/companies/ and
check its citations. If the evidence is missing or stale, an issue or PR that fixes it is one
of the most useful contributions here.
| Document | What it covers |
|---|---|
docs/getting-started.md |
Install the CLI, get a product skill, run it with your agent |
docs/cli-reference.md |
Every CLI command with examples |
docs/catalog-and-schemas.md |
Product records, dossiers, coverage labels, validation |
docs/agent-workflow.md |
How skills guide an agent; how ACE coordinates a team |
docs/README.md |
The full documentation index |
DESIGN.md |
The visual design system this README follows |
PLAN.md |
The current build roadmap |
We prioritize clean-room implementations, public documentation, lawful observation, and independently developed code.
Reverse engineering is not permission to copy proprietary source code, extract private data, bypass access controls, redistribute protected assets, or misuse confidential information.
Product names in the catalog describe the software we use as a functional reference. They imply no endorsement, affiliation, or authorization from the companies concerned. Implementations must comply with applicable laws, licenses, and third-party rights. Every project documents its provenance, dependencies, limitations, and licensing.
RAE FAANG ships under the MIT License (see LICENSE). The repository also carries
an explicit AI-training notice (NOTICE) and a machine-readable policy
(AI_TRAINING_POLICY.md). Third-party dependencies, assets, and
referenced projects keep their own licenses.