Browse docs
BetterWright documentation.
One persistent, policy-guarded browser engineered so every observation costs the fewest tokens possible. Used two ways — integrated, where your agent drives, or standalone, where BetterWright’s own agent loop drives and returns one JSON answer.
Start here
- Getting startedinit, the two usage shapes, first run, sessions, proof screenshots
- Integration guideWiring BetterWright into any host — written to be followed by an agent
- The built-in agentbetterwright exec, the interactive console, model adapters, runAgentTask()
- Sessions & the daemonPersistence, concurrency, interrupting a run, reconnecting
API
Capabilities
- Credential vaultEncrypted storage, site matching, selector-free login, betterwright vault
- Live view & handoffWatch, coach, or take over in a browser tab; hosting and security model
- CAPTCHA helpersLocal checkbox/Turnstile/slider solving; vision handoff for image grids
- Skill packsPer-site guidance the host agent reads on demand
- Network policyWhat the browser may reach; the unliftable metadata floor
- Agent guidanceThe operator prompt and its guardrail options
Under the hood
- Architecture & securityThe worker process, the RPC loop, what is and isn't a security boundary
- Cloaking V2The coherent-fingerprint approach; launch modes and egress matching
- Chromium forkBetterWright's own build: farbling, platform masking, discovery
- Fork patchesWhat each patch in the Chromium fork changes and why
- Headed / headlessDisplay modes over one persistent profile
- Security policyThe trust model, reporting a vulnerability, the shell as a trusted channel