WebCLI - Agent Interface for the Web.

Stop doing web chores your agents can do instead.

Productivity

Developer Tools

Artificial Intelligence

Visit WebsiteSee on Product HuntTwitter ⧉

Hey Product Hunt — I’m Cris, the builder who made WebCLI. I made WebCLI because agents got lost and stuck using browsers for tasks on the web. MCP, screenshots are slow, fragile and too token hungry for all interactivity; automation scripts are too fragile and static, and agents just ove CLIs. Install WebCLI - then tell your agent a goal, and with the WebCLI skill loaded it can figure it out on the web, on its own. The only time it stops for your help - is when it needs logins, captchas, sensitive approval, or when it just gets confused. It has saved me from annoying dashboard tasks (like DNS, flight booking, VM launching, etc) so I can just focus on stuff I want to do. A few examples of prompts you can give an agent with WebCLI: “Use WebCLI to compare prices across these vendor portals and give me a report. Then buy the best one.” “Use WebCLI to check AWS, GCP, and Azure for the cheapest safe VM option then launch it.” “Use WebCLI to setup this domain setup from Namecheap to Cloudflare, then deploy the Pages site.” “Use WebCLI to inspect this SaaS dashboard, find the lagging tasks, and send reminders to assignees to keep things moving." “Use WebCLI to find a flight and hotel for these dates, compare options, and book the best one. Fill in my details, I'll do payment.” WebCLI translates the web into agent-native language and gives your agents an interface they already know and love - a well-designed command line tool. WebCLI is free for 5 days per email domain, then $5 or various tiers. The main way to interact with the web until now has favored humans or bots. Humans love a GUI, bots need an API like Puppeteer, but agents need a TAI for the web - a Textual Agent Interface. WebCLI is that interface. WebCLI lets your agents drive browser tasks from their native environment - the command line - while you handle logins and captchas. No magic, just agent-native browsing unlocking tasks on the web. The bigger backstory is WebCLI exists because AI agents can code, but a most work lives in the web - especially the annoying choregraphy of multiple tabs and sites just to get shit done. Portals, forms, frames, settings, consoles, tools, checkouts, booking, vendors, accounts, etc etc etc. WebCLI gives agents the browsing skills they can actually use to accomplish anything you throw at them. It's not a new browser - it works with Chrome and FF - it's a new interface to browsers. WebCLI is a local-first CLI + agent skill. The CLI gives your agent control. The skill teaches it how to drive. observe → choose → act That's the core OODA loop. And you jump in to help if the agent gets blocked (on login gates, captchas or just confused). No magic, just your help when your agents needs it. It's not a service, it's a downloadable software that respects you. It's not another way to make brittle browser scripts, or static RPA -WebCLI deals with the real web as it is, not how a bot wishes it would be. WebCLI is for web work that humans, now agents can do it. Your agent/model is the intelligence, WebCLI is the precision optics, and you set the goals. WebCLI lets AIs see what they're doing without heavy screenshots and act precisely without brittle selecotrs, or MCP. You just tell your agent the goal, they decide the next move, figure shit out, and only ask you for your human-help when needed. When the web needs a human — login, MFA, CAPTCHA, payment, authorization, escalation — WebCLI does not pretend it can do it. It gets your agent to stop, explains the blocker, and hands off to you to help. That's all you do, your agent with WebCLI does the rest. It's local-first. No telemetry. No training on your data. No selling your info. WebCLI lets agents improvise, because that beats scripts and bots for real tasks, every time. I’d love feedback from builders, and AI users. Try it here: https://webcli.sh