Docs

Every doc lives in the repo. Each link goes to the doc on GitHub. Pages on this site have a Markdown copy: add .md to the address.

Start here

Install channels: every way to install webular besides npm.
How webular works: each building block and what it does for you.
Why webular: where webular differs from hosted rivals, and where they do better.
Benchmarks: scores against other web research tools, losses included.
Benchmarks by peer: webular against each peer on the tasks both answered.
How to read the benchmarks: what was measured, why, and every term in plain words.
How the benchmarks are run: the method, how to run it again, the raw data rules, and outside scores.
Command parity: each command’s rival endpoints, whether the bench scores it, and the metric that fits.

Command reference: what each command does.
Agent contract: what an agent needs from the CLI, MCP server, plugin, and skills.
Browser mode (Cua): how webular drives a throwaway browser window.
Reliability: retries, dead pages, and sites that will not answer.
Research factors: each factor that makes web research correct, fast, cheap, and repeatable, with its guard or its gap.

Architecture: how a command runs and which XState machines model it.
Development: read before you change webular or run its tests.
Build and run: run from source, use the bundles, or build the executable.
Bug classes: each kind of bug webular can hit, where it shows up, and what guards against it.
Mise tasks: every mise task, what it does, and what it runs first.
Best practices: each practice webular follows, with its source and check.
Research ledger: the papers behind webular’s design choices, and the ranked changes each topic suggests.
Research status: how many papers and repos were read, and which ranked changes are applied, rejected, or open.
Jev in webular: how to set up the Jev key on any OS, and each route webular takes to Jev for its judgment calls.
Docs rules: the rules for writing these pages.

For agents: llms.txt indexes every page.