BuildSeniorAI engineeringAuto-graded
Build a tool-using agent with Claude Code
An open-ended build that reveals whether someone can get AI tooling to finish real multi-step work unsupervised.An open build rather than a repair. You have Claude Code and a set of tools — file I/O, web fetch, shell — and a multi-step research task to finish end to end. Nobody will tell you the shape of the solution: you decide how to decompose the work, which tools each step needs, and how the agent knows it is finished. The deliverable is a working agent plus the report it produces, with its sources and its own confidence stated.50 min
python
You're done when
- An agent is actually implemented, not just sketched or described.
- It drives a real model through the Anthropic or Claude Code interface rather than faking the calls.
- It declares the tools it can use, so the model knows what is available to it.
- It runs a genuine multi-step loop instead of a single request and response.
- It writes out a report as its final artefact.
- That report carries a summary, the sources it relied on, and a stated confidence level.
What it measures
Agent engineeringBuilding a loop that completes multi-step work unsupervised, including deciding when it is finished.
Claude CodeGetting real work out of agentic tooling on an open-ended task, rather than using it as autocomplete.
Context engineeringChoosing what the model sees at each step and which tools to expose, so the agent can act without being flooded or starved.
Your environment
Terminal + editorA full shell in the browser
Claude CodePre-authenticated, no key setup
Isolated microVMYours alone, wiped afterwards
Start this lab
50 min
Start labYou'll sign in to Anthropos to start. The lab opens in a private cloud workspace with Claude Code ready to use.AI Labs is part of Anthropos Workforce. Not on your plan yet? Book a demo