AgentR today launched Webcmd, free open-source browser infrastructure that lets AI agents learn a website once and reuse that knowledge on later runs, passing the most tasks at the lowest cost per task in AgentR’s published BU Bench V1 comparison.
— San Francisco Bay Area, October, 8, 2026 — AgentR today launched Webcmd, free open-source browser infrastructure that lets AI agents learn a website once and reuse that knowledge on later runs instead of relearning the site each time. Webcmd is maintained by Nishant Gaurav and his team at AgentR, and is available now on GitHub under the Apache 2.0 license.
Browser agents without memory start from zero on every visit. They re-read pages, hunt for the same controls and repeat the same steps, and each of those steps costs tokens. Webcmd records what an agent observes on a site, including pages, states, actions, workflows, APIs, pitfalls and fallback paths, as an agent-facing sitemap stored locally. The next agent to visit that site starts from the map. AgentR’s stated goal is to cut browser-agent token spend by up to 90%.
Photo courtesy of AgentR
On BU Bench V1, a 100-task browser automation benchmark published by the browser-use project, Webcmd passed 67 tasks, the highest score in AgentR’s published comparison. Its estimated cost was $0.255 per completed task, the lowest of the five tools tested, against $0.297 for browser-use and $0.554 for agent-browser. Webcmd averaged 9.8 agent turns per completed task, 34% fewer than browser-use at 14.8.
“We are building systems that learn from real use, not just from training data,” said Gaurav. “That is the shift we want people to understand, because it changes what software can become in practice.”
Webcmd lets an agent send one sandboxed, Playwright-style JavaScript program that chains related browser steps, rather than issuing one command at a time. It trims page snapshots to the content an agent needs and returns only what changed after an action. The live browser is always treated as the source of truth, and a memory failure never blocks a task.
Developers found the project early. On Aug. 22, Webcmd was the No. 1 trending JavaScript repository and No. 4 overall on Trendshift, which tracks GitHub activity. It has passed 2,600 stars and 1,600 forks, and has been shared in Reddit communities for Claude skills, the Hermes agent and the OpenCode CLI.
The launch comes as the cost of running agents draws scrutiny. Gartner predicted in June 2025 that more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls.
AgentR has published the full Webcmd benchmark run and reproduction steps on GitHub. Webcmd installs with a single npm command, npm install -g @agentrhq/webcmd, and requires Node.js 20.6 or later.
About AgentR
AgentR builds self-learning agentic AI systems, software designed to improve through use rather than stay fixed after release. Founded in late 2025 and backed by investors including General Catalyst, the company is led by Nishant Gaurav. Its products include Webcmd, a self-learning agent browser that helps AI agents remember the websites they use, and Authsome. AgentR releases Webcmd as open source so developers can inspect, run and reproduce its results on their own machines. The Webcmd repository, maintained by Gaurav and the AgentR team, has nine contributors and is licensed under Apache 2.0. Learn more at webcmd.dev and github.com/agentrhq/webcmd.
Media contact
Nishant Gaurav
Co-Founder, AgentR
nishant@agentr.dev
+91-90045-15020
Contact Info:
Name: Nishant Gaurav
Email: Send Email
Organization: AgentR
Website: https://agentr.dev/
Release ID: 89205663
If you detect any issues, problems, or errors in this press release content, kindly contact error@releasecontact.com to notify us (it is important to note that this email is the authorized channel for such matters, sending multiple emails to multiple addresses does not necessarily help expedite your request). We will respond and rectify the situation in the next 8 hours.