Using it

Start here if you want to run Browsentic on your own machine.


Install	Prerequisites, clone, build, load the extension
Pair	Put the CLI on your `PATH` and connect your browser
First run	A tour of the side panel and your first instruction
Features	One page per capability — what it does and when to reach for it
Choosing an agent	Claude Code, Codex or Antigravity
MCP clients	Drive the same browser from Claude Code, Cursor, Zed, Gemini CLI
Configuration	Every key in `~/.browsentic/config.json`
Approvals	What asks before it acts, and how to change that
Limits	Where Browsentic does not fit — read before you rely on it
Troubleshooting	Symptom → cause → fix
Maintenance	Updating and uninstalling

Building on it

How the pieces actually work, for contributors and for anyone integrating.


Overview	Four processes, and why there is a daemon at all
Transport	Ports, the origin gate, the pairing handshake
The action registry	One definition, two bundles, and drift detection
Request path	An MCP tool call, end to end
Inside the extension	Background vs content script, tab scoping
Agent runs	The intent funnel, runners, prompt assembly
Guardrails	The policy, run scope, fencing, spawn containment
Subsystems	Monitors, recordings, site maps, files, screenshots
State on disk	What is written where, and at what mode
Contributing	Build topology, checks, adding a capability


Tools	All 41 MCP tools with their parameters, plus the resources
CLI	Every `browsentic-mcp` command
Errors	Every error code, what caused it, what to do

New here? Install → Pair → First run.

The project README is the two-minute version of all of it.