Home stack
Claude
Build with Claude this week. Install in order. One lab per step. One artifact. This site is independent and not affiliated with Anthropic — Claude is the home stack on this domain. Cursor and Windsurf stay on Starting a project.
01Install Claude Code
Agents / harnessThe coding agent on this company’s repo this week. Not a demo folder.
Artifact One real PR or ticket Claude Code helped close. · board Agents / harness
02Wrap Claude in the app
LLMs / contextCall the Messages API from the product. The model is a dependency, not the product.
Lab Get started · Working with messages
Artifact A Messages call in THIS app, with a model you can swap. · board LLMs / context
03Put the right things in the window
LLMs / contextContext is what Claude can see this turn. Assemble it for THIS repo.
Lab Effective context engineering for AI agents
Artifact A reusable context / prompt assembly checked into the repo. · board LLMs / context
04Give it hands, then bound them
Agents / harnessTools, permissions, MCP. The wrapper is the product.
Lab Effective harnesses for long-running agents · Code execution with MCP
Artifact Bounded tools + permissions on that project. · board Agents / harness
05Tool use — computer use only if the UI is the API
Agents / harnessSkip computer use unless the product must click, type, or look at a screen.
Declare tools Claude can call. Computer use is for a real desktop or browser, not a chat demo.
Lab Tool use · Computer use
Artifact One tool Claude can call. Computer use only if you needed the skip-off. · board Agents / harness
06Ship a Claude app
ProductionPut the app on cloud with Claude as the model. Not a VPC. Not a training cluster.
Lab AI SDK Anthropic provider · Next.js App Router + AI SDK
Artifact A URL. · board Production
Need the vendor-neutral Monday order? Starting a project. Heard the slang first? New here. The board is still the firehose. Independent — not a product of Anthropic.