Reading up on WebMCP
11 deep Β· digging since mar 26
- Automating repetitive work at OpenAI with Codex
The author describes using Codex with Runme notebooks and WebMCP to automate repetitive engineering tasks at OpenAI, turning evaluations into reviewable, reusable workflows.
- π¨ Breaking π¨ ChatGPT Now Supports WebMCP - by nekuda
OpenAI announced WebMCP support in ChatGPTβs desktop browser and Sites, enabling agents to use websiteβexposed tools directly for faster, reliable web tasks.
- The State of WebMCP: July 2026
WebMCP is a live Chrome origin trial standard with nearβzero adoption and no AI agent support yet, but site owners can prepare by adding tools and monitoring Lighthouse audits.
- Lighthouse agentic browsing scoring
Lighthouse's Agentic Browsing category scores sites as a pass/fail ratio for machine-readiness, using audits for WebMCP, accessibility tree integrity, stability, and llms.txt.
- Where my head is at with LLMs and the web
LLMs lower barriers to the web but inherit a stale, outdated model of the platform; new primitives like WebMCP and agent loops could reshape the web's future.
- Browser Run: give your agents a browser
Cloudflare renames Browser Rendering to Browser Run, adding Live View, Human in the Loop, CDP access, WebMCP support, session recordings, and 4x higher concurrency for AI agents.
Takes
π I made a small Chrome extension: a side panel that drives any site's WebMCP tools with Jev! When you type, on every keystroke it picks the relevant page's tool, fills in the arguments, and tells you how sure it is. Here it is grocery shopping:
@sarah_edo
I've been working on this app to keep myself organized, but WebMCP makes it pretty killer. π₯ See how I can seamlessly create a workback gantt chart just from some board items?
@sarah_edo
Introducing WebMCP + agent-browser The web is becoming agent-native πππππ-ππππ πππ π πππππ <ππππ|ππππππ> β Discover & call tools β For missing tools, falls back to browser automation
@ctatedev
Really interesting new blog post from @openai for several reasons: 1) Shows an example of building with WebMCP, meant for when you want agents and and humans to collaborate on using a UI (like co-editing notebook cells). It's different than MCPs or APIs in that its exposed directly through the browser. Read the post for discussion of the tradeoffs. 2) They created a new kind of notebook which works with WebMCP that prioritizes meeting people where they are: you bring your own coding agent and files are just markdown. The author uses it to curate runbooks or high quality examples of how to run foundation model evals on their infrastructure. Notebooks are good for this since they require tinkering with state of long running jobs interactively while taking notes inline. And its open source β¨ Blog:
@HamelHusain
OMG you guys, this is incredible! This is using Google's new WebMCP function to control your browser, but not only is it lightning fast, but its unique because it is using your main Chrome instance. Not some sandboxxed Playwright instance that doesn't want to remember your sessions, cookies, or passwords. Your real Chrome instance. It's incredible. You need to enable: chrome://inspect/#remote-debugging Also, it doesn't even require a skill to use. It just works. I'm thinking about making one anyway. I'm telling you, download this and try it. This is my new daily for sure.
@LLMJunky