The built-in tool catalogue, and what it costs
You do not have to write everything. Anthropic ships a catalogue of tools it defines for you, and Agentic Customization is 4.1% of the paper, which is mostly this list plus knowing what it costs.
Two axes run through the catalogue and they are independent. GA versus beta is about how settled the interface is. Server versus client is about who executes the call. Sort the list by one axis and the other stops being visible, which is where the marks go.
The GA list, by who runs it
Anthropic executes these. No handler code in your application at all:
Code execution. Runs code in a sandbox. GA, despite being the one that sounds like it should still be experimental.
Tool search. For when you have too many tools to put all their definitions in every request, which is a real problem the moment you connect a couple of MCP servers.
Web search. $10 per 1,000 searches, plus the tokens the results burn.
Web fetch. Free, plus tokens.
Anthropic publishes the schema and trains the model on it, and you execute these:
Bash. Shell commands. Claude emits the command; your process runs it.
Text editor. Structured file viewing and editing, the same shape of capability Claude Code uses. Your process touches the filesystem.
Memory. Persistence across turns without you hand-rolling the schema. You still supply the store.
Computer use is the beta
Computer use is still beta. Beta header computer-use-2025-11-24, and computer-use-2025-01-24 for older models.
Everything else on the list above is GA, and the intuition that “the scary one must be the beta one” leads you straight to code execution, which is wrong. Release status tracks how settled the interface is, not how dangerous the tool sounds.
The pricing shape that changes designs
Look at search and fetch together, because the gap between them is a design instruction.
Search costs money per invocation. Fetch does not. Which means: discovery is billed, retrieval is not. An agent that already knows the URL and pulls it is free at the tool layer. An agent that searches the open web on every turn because it never bothered to remember anything has a line item.
That is not a micro-optimisation, and the reason is the loop. Everything inside an agent loop gets multiplied by turn count, and turn count is the number nobody bounds until the invoice arrives. A habit that costs a fraction of a penny per turn is invisible in a demo and a budget line in production. Caching a URL is not premature optimisation here; it is the difference between two pricing models.
The versioned type strings
Every built-in tool is referenced by a dated type string, and those dates move. You will see this concretely in the next lesson, where code_execution_20250825, code_execution_20260120 and code_execution_20260521 all exist, all listed GA, and the documentation is not consistent about which one a given feature wants.
Do not try to memorise the current date for every tool. Memorise that the dates exist, that several versions can be live simultaneously, and that this is a lookup on exam day. Knowing which facts are lookups is worth marks on a timed paper, because it stops you burning three minutes trying to recall something you were never going to recall.
Try it yourself
The odd one out
Six of these are GA. One is not, and it is the one people assume is furthest along because it demos the best.
Show answer
Correct answer: C — Computer use
Computer use is still beta and carries the header computer-use-2025-11-24, with computer-use-2025-01-24 for older models. Code execution is the tempting pick because it sounds like the riskiest thing on the list, running arbitrary code, so surely that is the cautious one. It is GA. Riskiness and release status are not correlated here.
Search and fetch, priced
Two web tools, two very different bills. Exact numbers, because the exam quotes exact numbers.
What does web search cost, and what does web fetch cost?
Reveal answer
Web search is $10 per 1,000 searches, on top of the tokens the results consume. Web fetch is free, again on top of tokens. So the expensive verb is search, the discovery step, and pulling a URL you already know is not billed as a tool at all. An agent that searches on every turn out of habit has a real line item attached to that habit.
Price the habit
Your agent runs 20 turns per session and calls web search on 8 of them. You run 500 sessions. What does the search tool cost you, tokens aside?
Show answer
Correct answer: C — $40, because that is 4,000 searches at $10 per 1,000
8 searches times 500 sessions is 4,000 searches, and at $10 per 1,000 searches that is $40, on top of whatever the results cost in tokens. The per-turn answer is the designed trap, because turn count is the number engineers actually track and it is the number that feels like the unit of work, but the billing unit is a search and twelve of those twenty turns never made one. The free answer is web fetch's story, not web search's, and mixing the two up is the single most useful confusion to clear before the exam.
Two headers, two model generations
This is the kind of fact that is pure lookup in real life and pure recall on an exam.
Show answer
Correct answer: A — computer-use-2025-11-24 for current models, computer-use-2025-01-24 for older ones
The current header is computer-use-2025-11-24 and the older-model header is computer-use-2025-01-24. The reversed answer is the trap and it catches people who remember there are two dates but not which way round they go: the later date is the newer capability, which is the direction it always runs. The no-header answer fails on the prior fact that computer use is still beta.
Two minutes, docs open, which version
You need the current type string for the code execution tool, you know several versions exist, and you have not memorised any of them.
Show answer
Correct answer: D — The tool reference for code execution, which lists the versioned type strings
Versioned type strings live on the tool reference, and this is exactly the sort of fact worth looking up rather than recalling, because more than one version is live at the same time. The beta-headers answer is the tempting one, and it is the same inference the next lesson takes apart: the Agent Skills header is dated 2025-08-25 and one tool version reads 20250825, so it feels as though one tells you the other. Beta headers and tool type versions are separate axes that happen to share a date. Pricing does not vary by tool version, and a changelog tells you what changed rather than what is current.
Audit your own agent
Take the research assistant from the capstone brief, or any agent you have actually built, and list every built-in tool it would use.
For each tool: is it GA or beta, does it need a beta header, and what does a thousand invocations cost you. If you cannot fill the cost column for web search without looking it up, that is the card to review tonight.