From the course: AI Trends
Claude Code on the web
- Claude Code has been a really important AI tool because developers rely on it so much for agentic coding. It has a well-respected CLI and extensions for your favorite IDE, but it's been missing one thing, and that's the web. The new Claude Code on the web lets you run, review, and collaborate on code from anywhere. When you go to claude.ai, you can find Claude Code sitting right there underneath your other tools, and that's one of the huge advantages of this tool. There's no installation, local setup, or extensions. Just start chatting with your repos directly through the web. You will see this Install GitHub App option here, and that's gonna let you connect this to any repositories that you want to. From this install screen, you can choose any one of your organizations. You can give it access to all your repositories or only search and repositories. Once you do that, you'll see all of your repositories available right here. You can do a search for which one you want to work with, and you get a dropdown so that you can see the list of active or archived repositories. You usually keep this pretty clear. And this interface lets you control one or more repositories with Claude. You can ask questions, have the tool review some code, or create a new feature, and it can easily work on multiple tasks concurrently. Let's start by asking it a generic question about this repository. I'll say, "Tell me about this repository," and then I'll hit Send and it's gonna create a session and give you the results of the prompt. All the tasks are going to run a synchronously on Anthropic-managed virtual machines, which have their own environments that you can customize and you can manage them right here so you can work with the default environment and control what sort of access you allow for this session. Trusted network access means that you're allowing it to work with whitelisted domains. You can take a look at more information on the benefits of sandboxing. This is really a way to keep things a lot safer. Notice that you can also add variables directly onto your environment, so if you're using any API keys, you can add them in here. You can also, of course, create your own environment and it gives you those same options. Let's try making a change. I've got this project loaded locally. This project is a dashboard for podcast information. In this episode duration versus completion section, I've got some popovers that let me look over the individual dots, but that doesn't work on some of the other ones, so I'm gonna add a prompt to fix that here. I'll say, "In the episode duration and completion section, I've got some popovers that let me move over the data points to see the values. Can you implement that in other graphs?" And it makes a to-do list to accomplish the tasks, and then it starts working on taking care of those issues. When it's done, it's going to give you a report of everything it's done and it's going to activate the options to open in CLI and create PR. When you click on Open in CLI, it's going to copy a command in the clipboard and you can paste that in your terminal, which is going to bring up that project inside Claude. This tool allows you to create a pull request, so I'm gonna create PR right here, and now it takes me to GitHub where I can review the pull request. I'll go ahead and create it. It's gonna run the usual checks, and then I can merge this pull request. Now you can see that same functionality has been added everywhere in our project. All right, so why would you want this? Well, the extensions and the CLI are great ways to work with individual projects, but this can be a great solution for managing your entire codebase.
Contents
-
-
-
Microsoft Ignite8m 23s
-
Google Gemini 39m 52s
-
NVIDIA DGX Spark2m 10s
-
Kimi K2: The thinking open-source model6m 59s
-
Agentic development changes with Cursor 24m 39s
-
Agentic browsers5m 12s
-
Claude Code on the web3m 33s
-
Claude Skills6m 10s
-
(Locked)
OpenAI Agent Builder6m 43s
-
(Locked)
OpenAI Apps SDK6m 45s
-
(Locked)
GitHub MCP Registry6m 22s
-
(Locked)
GPT-5-Codex: A coding paradigm shift3m 41s
-
(Locked)
MCP in ChatGPT Developer Mode7m 23s
-
(Locked)
Styles and modes in ChatGPT, Claude, and VS Code6m 34s
-
(Locked)
Context Engineering3m 27s
-
(Locked)
Vibe coding with Windsurf6m
-
(Locked)
OpenAI GPT-511m 41s
-
(Locked)
OpenAI gpt-oss7m 11s
-
(Locked)
Perplexity Comet7m 34s
-
(Locked)
ChatGPT agent6m 56s
-
(Locked)
OpenAI's o3-pro model4m 32s
-
(Locked)
Gemini Diffusion model7m 47s
-
(Locked)
Codex54s
-
(Locked)
Bolt45s
-
(Locked)
Vibe Coding3m 39s
-
(Locked)
Generative Engine Optimization (GEO)4m 9s
-
(Locked)
Github Copilot Spaces5m 22s
-
(Locked)
Model Context Protocol (MCP)4m 32s
-
(Locked)
Physical AI2m 17s
-
(Locked)
ChatGPT 4.53m 25s
-
(Locked)
Coding in Cursor1m 15s
-
(Locked)
OpenAI's Operator3m 20s
-
(Locked)
Choosing an OpenAI model5m 9s
-
(Locked)
DeepSeek7m 20s
-
(Locked)
OpenAI canvas4m 58s
-
(Locked)
Agentic computer use from Claude1m 58s
-
(Locked)
NPUs vs. GPUs vs. CPUs2m 45s
-
-
-