toolcall()

// AI dev tooling, explained fast.

Stop guessing how it works.

One AI dev-tooling concept — MCP, agents, RAG, prompt caching — properly explained every week. The mechanism, not the hype.

scroll

From docs to done.

The docs are a maze and the threads are noise. ToolCall takes one concept a week and draws the line straight through it — so you actually understand the tool before you ship with it.

Learn fast.

30–60 seconds. One concept, one mental model, no preamble.

Skip the trap.

The gotcha that bites in production — before it bites you.

Keep the good ones.

Save the breakdowns worth re-reading. A reference you build, one a week.

We don't decorate.
We explain.

Every concept gets the same treatment: name it, draw it, show the one rule that breaks it. Checked by someone who builds with these tools — no generic AI filler.

What we explain

all concepts →
01

MCP & tools

How models actually reach your data and systems — the M×N → M+N problem it solves, the three primitives, the transports.

read →
02

Agents

The loop behind the buzzword: think, act, observe, repeat — and where it quietly goes wrong.

soon
03

Cost & caching

Prompt caching, token math, and the structure choices that cut cached-input cost by ~10×.

soon
04

RAG & context

Retrieval, chunking, context engineering — what to put in the window and what to leave out.

soon

How each one is made

// the method
01 / pick

One concept

The thing devs keep getting wrong this week.

02 / verify

Owner-checked

Every claim signed off by a practitioner.

03 / draw

Built to show

The mechanism animated, not narrated over stock.

04 / ship

Save & reuse

60 seconds you can act on, then keep.

Get MCP in one page.
Free.

The diagram, the three primitives, a runnable starter — then one concept a week.

    // total: 0.00 · spam: void · unsubscribe: one click