# Bowmark > The web as a typed capability library. An agent reads the library, writes a short JavaScript script against it, and Bowmark runs that script on the live sites. ## Docs - [HTTP API](https://bowmark.ai/docs/api.md): The base URL, the auth header, and the endpoints — for a caller with no MCP client and nothing installed. - [Browser Agent](https://bowmark.ai/docs/browser-agent.md): The paid last-resort fallback for a page that needs clicking, a date picked, or a login walked through — the real start/status/send/stop shapes and the time limits. - [Evals](https://bowmark.ai/docs/evals.md): Real recorded runs — the prompt as a person typed it, the script the model wrote, and the answer that came back off the live sites. - [How it works](https://bowmark.ai/docs/how-it-works.md): Two calls. Read the library, then run a script against it on the live sites. - [Installation](https://bowmark.ai/docs/installation.md): Point an MCP client at https://api.bowmark.ai/mcp with your API key, or install the client library for your own code. - [LangChain](https://bowmark.ai/docs/langchain.md): Give a LangChain agent Bowmark's typed functions for live websites with the langchain-bowmark package — two tools, one API key, structured results. - [Pricing](https://bowmark.ai/docs/pricing.md): How Bowmark charges. US dollars for the resources each call uses, $10 free every month, a card on file for the rest. - [Providers](https://bowmark.ai/docs/providers.md): Check whether Bowmark covers a site. Agents and fetch tools: call GET https://api.bowmark.ai/v1/library?query=, or read PROVIDERS-SIMPLE.txt — do not summarize this page. - [Quickstart](https://bowmark.ai/docs/quickstart.md): Get one real result out of Bowmark before you install anything, then wire it into your agent. - [Scripting](https://bowmark.ai/docs/scripting.md): Write against the library from your own Node or Python program, or as a script string inside an agent. Both surfaces, and when each is right. Full text of every page: https://bowmark.ai/llms-full.txt