What Xymotrone can do
One agent loop, real browsers, your network and your logins — with the operational controls needed to run it more than once.
The agent
Read, decide, act
Each step the model receives the page, picks a single action — click, type, key press, scroll, tab switch, navigate — and Xymotrone executes it in real Chrome.
Any model you like
Hosted models out of the box, or point Xymotrone at your own Ollama, LM Studio or vLLM endpoint and test it from the setup wizard before use.
Screenshots and reasoning
Every step stores the thought, the action, the observation and a screenshot, so you can audit or replay any run afterwards.
Scale and network
Parallel sessions
Launch several instances of the same task, each in its own browser with its own inputs, all monitored from one live panel.
Proxy pools and health
Group HTTP, HTTPS and SOCKS5 endpoints into pools with round-robin or random rotation, and check exit IP, latency and failures before a run starts.
Fingerprint controls
Per-session fingerprints with WebGL spoofing and canvas/audio jitter, plus a clean session wipe when the run ends.
Operations
Credentials bank
Store logins once, match them to a site automatically, and rotate through a pool so no two parallel sessions reuse the same account.
Scheduling
Give a task its own date and time, or repeat it on an interval — scheduled runs go through the same limits and logging as manual ones.
Timeouts and retries
Per-step timeouts, step retries, a maximum step count and a run deadline keep a stuck agent from burning your quota.
Ready when you are
Every plan includes all of the above — plans differ only in run volume and how many browsers work at the same time.