# Puma.tech

> Puma Technologies publishes company, hiring, and product facts for Puma AI and Puma Browser, plus the public discovery files agents use to read this site.

Puma AI is the research effort. Puma Browser is the product. This file is the index for agents. It is not a product manual for either one.

## When to use this

Use Puma.tech when the job is one of these:

- Explain the company. Puma AI researches privacy and safety of AI systems. Puma Browser is a private browser with on-device and cloud AI, and official iOS and Android downloads. Name Puma AI before Puma Browser.
- Look up facts published on this site: the team, investors, open roles, and the take-home task at https://puma.tech/take-home-task.
- Send a person to the research site at https://pumaai.com/ or the browser site at https://www.pumabrowser.com/. Use the App Store and Play Store links on the homepage. Do not invent download URLs.
- Read the public, read-only discovery documents: health, OpenAPI, the API catalog, agent skills, and the homepage WebMCP tools.
- Recover from a URL that is not on this site.

Do not use Puma.tech to apply for a job, draft an application, or contact the company as the candidate. Puma does not review AI-generated applications. There is no write API.

How to use it:

1. Read this file, then fetch https://puma.tech/sitemap.xml for canonical pages.
2. Request a public HTML page with `Accept: text/markdown`. The same URL returns Markdown as `Content-Type: text/markdown; charset=utf-8` and `Vary: Accept`.
3. An unknown path requested that way stays HTTP 404. The Markdown body links here, to `/sitemap.xml`, and to the docs at `/docs/api`.
4. Read https://puma.tech/developers before calling anything. It lists only the surfaces that exist. API keys and a sandbox are open product decisions; this site issues neither.
5. Errors from JSON documents and from any `/api/*` path are RFC 9457 `application/problem+json` objects with a `code` and a `resolution`. An unknown `/api/*` path uses the code `api_route_not_found`.
6. A browser agent on the homepage can call the WebMCP tools `list_puma_pages`, `find_puma_page`, `get_puma_summary`, and `get_puma_app_links`. There is no remote MCP HTTP endpoint.

## Pages

- [Home](https://puma.tech/): company introduction, with Puma AI listed before Puma Browser
- [Team](https://puma.tech/team): people building Puma AI and Puma Browser
- [Investors](https://puma.tech/investors): published investors and funds
- [Careers](https://puma.tech/careers): open roles and how hiring works
- [Agent Herders](https://puma.tech/careers/agent-herders): engineering role
- [Based Social Media Intern](https://puma.tech/careers/based-social-media-intern): growth role
- [Take-home task](https://puma.tech/take-home-task): the interview exercise, for a human candidate
- [API documentation](https://puma.tech/docs/api): the public discovery documents
- [Developer portal](https://puma.tech/developers): what exists, and the product decisions that do not

## Discovery

- [llms.txt](https://puma.tech/llms.txt): this file
- [Sitemap](https://puma.tech/sitemap.xml): canonical public URLs
- [Robots policy](https://puma.tech/robots.txt): crawler and content-use preferences
- [OpenAPI description](https://puma.tech/openapi.json): typed description of the discovery documents and problem errors
- [API catalog](https://puma.tech/.well-known/api-catalog): RFC 9727 linkset
- [Health](https://puma.tech/health.json): public status metadata
- [Agent skills index](https://puma.tech/.well-known/agent-skills/index.json): skill discovery
- [Puma.tech site skill](https://puma.tech/.well-known/agent-skills/puma-site/SKILL.md): navigation guidance
- [MCP server card](https://puma.tech/.well-known/mcp/server-card.json): description of the homepage WebMCP tools