Agents

Claude Code, Codex and your own scripts can drive Eimyr: open pages, read them, click and type, look at the console and network. They work in a Space of their own, with its own profile, signed in nowhere you are, and you watch every step.

For Eimyr's current version

Connect an agent

Eimyr's MCP server comes inside the app. Add it to your agent once:

claude mcp add eimyr -- /Applications/Eimyr.app/Contents/Helpers/eimyr mcp
codex mcp add eimyr -- /Applications/Eimyr.app/Contents/Helpers/eimyr mcp

Any other MCP client takes the same command: eimyr mcp, on standard input and output. The first time an agent connects, Eimyr asks you whether to let it.

Scripts can call the same tools as commands: eimyr open example.com, eimyr snapshot t1 --interactive, eimyr click t1 e12. eimyr guide prints everything below with every command.

How an agent works in Eimyr

What stays yours

Every tool

open_tab, list_tabs, navigate, go_back, go_forward, reload, snapshot, find, inspect, click, hover, drag, type, select_option, press_key, scroll, get_text, describe_page, wait_for, screenshot, save_pdf, console, evaluate, emulate, handle_dialog, media, upload_file, sign_in, network, downloads, storage, ask_person, list_scenarios, run_scenario, batch, close_tab.

For agents reading this site: llms.txt.