Portelon
Portelon is a browser, for humans and agents.
It lets agents quickly, ergonomically, and token-efficiently do things a human would be able to do with a browser, such as researching and interacting with webpages, reading and filling out PDFs, and more.
Portelon works with any agent that supports MCP. It can be used with Codex, Claude Code, Cursor, GitHub Copilot, OpenCode, your own bespoke agent, and more. Installation and complete uninstallation are one-click for popular harnesses and straightforward for others.
Portelon is state-of-the-art on all tested benchmarks, spanning research and action.
Portelon makes a perfectly good daily browser, with built-in ad and tracker blocking. Agents share one profile with you, so they can use your own accounts and act on your behalf. Agents can also use your own tabs and yield control of their tabs to you when they need your help.
Humans and multiple agents can use the browser at the same time without disrupting one another, and by default, agents’ tabs are invisible and out of your way, unless they need your help.
The entirety of Portelon’s tool namespace amounts to around 280 tokens, which is a tiny fraction of the average context window—it isn’t bloated like many MCP servers. The main instructions, amounting to around 3,500 tokens, are loaded only when an agent chooses to use Portelon.
Portelon will rarely be blocked by bot-detection systems, because it presents the exact same fingerprint to websites regardless of who’s using it, and regardless of whether the browser is running invisibly or visibly.
Portelon runs privately on your computer, and there aren’t any artificial quotas or domain blacklists.
Portelon has a free trial.
Christopher Gait-Smith
Founder, Portelon, Inc.