Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

58 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

browser-memory

Docs npm version License MIT

Give your AI agent a reusable memory of web actions.


Install

Requires Node.js ≥ 20 and Google Chrome installed.

One command configures every supported agent it finds on the machine — Codex, Cursor, VS Code (Copilot) and Claude Code:

npx -y browser-memory install

Then restart the app: MCP servers are negotiated when the session starts, so a new chat isn't enough. It's idempotent (never overwrites an existing entry).

To install into a single host, name it (codex · cursor · vscode · claude):

npx -y browser-memory install cursor

Disconnect

npx -y browser-memory uninstall

…and restart the app. This is the only way to remove it — an MCP server cannot unload itself from a live session. It takes the same optional host as install, and it only touches the host's config: your learned tools (~/.tool-memory) and the Chrome profile are left alone.

Any other MCP host — drop this into its config (most clients use the mcpServers key; VS Code uses servers):

{ "mcpServers": { "browser-memory": { "command": "npx", "args": ["-y", "browser-memory"] } } }

Registry (optional)

On by default — it pulls ready-made tools from the hosted registry (https://api.browser-memory.com), anonymously: no account, no sign-up, no login prompt. If the registry refuses the request, you're told once and everything keeps working with your local tools.

Signing in is optional, and only if the backend asks for a key:

npx -y browser-memory login

(or set TOOL_MEMORY_REGISTRY_KEY yourself). The server never starts a login flow on its own.

Turn it off to run 100% local:

npx browser-memory config server off

Turn it back on:

npx browser-memory config server on

Point it at another backend:

npx browser-memory config set-url https://your-registry.example.com

About

Give your agent a reusable memory of web actions: learn a task once, then replay it deterministically.

Topics

Resources

Code of conduct

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages