Tue 09:12
You approve a 600-line diff you only skimmed, because reading it properly would take longer than writing it.
For developers who already work with Claude Code, Codex or Cursor
Agentic engineering, not vibe coding. Stop reading every line your agent writes.
AI Developer Toolkit is the manual for agentic engineering with Claude Code, Codex and Cursor. Its guides, Setup Pack and books show you how to hand over whole tasks and trust what comes back. You review the spec and the test report. The checks you set up review the diff.
$59.99 once for 12 months of access. Nothing renews on its own. 30-day refund on a first purchase.
- Guides for the three tools
- 400+
- Recipes you can run today
- 100+
- Book chapters, EPUB and PDF
- 328
- MCP servers in the Setup Pack, checked 2026-10-02
- 31
One task, start to finish, on a Level 4 setup.
Your attention goes to the spec before the build and the evidence after it. The diff goes to checks you set up once.
- Written by you
- Read by you
- Agent and checks
-
TicketIntent
PAY-412
Prorate plan changes made mid-cycle.
You write
-
SpecIntent
spec.md
Acceptance criteria, edge cases, files the agent must not touch.
38 lines
You write and read
-
PlanIntent
plan.md
Seven steps. The agent proposes them, you approve before it builds.
Agent writes
-
BuildLoop
+612 −148
14 files changed in an isolated worktree, until the tests pass.
Agent writes
-
VerifyVerification
- 212 tests passed
- Stop hook passed
- Review passed
760 changed lines read
Checks read
-
Pull requestRelease
#1184
Opened with the test report attached. You read it against the spec.
24 lines
You read
Harness: permissions, sandbox, MCP servers and hooks, installed once
Lines you read62
Lines the checks read760
- Drawing
- One task on a Level 4 setup
- Runs in
- Claude Code, Codex, Cursor
- Note
- An illustration of the workflow, not a measurement
Engineering teams at these companies use it
- Grupa Morizon-Gratka
- Ringier Axel Springer
- TasteRay
Creator of wondel.ai Skills — 2.3k+ stars on GitHub (opens in a new tab)
Your agent got faster. Your review queue didn’t.
Once the agent writes most of the code, you become its full-time reviewer. Every task ends in a diff you either read line by line or approve on trust. Approving on trust is vibe coding. Agentic engineering replaces that trust with checks that can say no.
“You are still responsible for your software, just as before.”
Andrej Karpathy, published transcript of his Sequoia Ascent talk (opens in a new tab), 2026-04-30
“Your life is diffs.”
Tue 11:30
Your CLAUDE.md started as a paste from someone’s thread. Nobody on the team knows which lines still do anything.
Tue 14:05
Claude Code, Codex and Cursor each shipped changes this week. You heard about them from a post on X.
From our reader survey
“I switch between Claude Code, Cursor, VS Code, GitHub, and other agents, and often have to explain the same project or problem again.”
A reader, July 2026
What two years of telemetry from 22,000 developers shows
Faros AI, “AI Engineering Report 2026”, April 2026 (opens in a new tab)
+66%
epics completed per developer
+242%
incidents per pull request
+441%
median time a pull request spends in review
Output went up, and so did incidents and review time. The way out is checks that read the code for you, set up once and run on every task. That is what the toolkit teaches.
Know your level. Then climb one rung.
Dan Shapiro’s five levels describe how much of the work you hand to an agent, from spicy autocomplete to the dark factory. The guides are filed by level, so the next step is always named.
Not sure which row is yours?
The AI Dev Score asks 25 questions in about 8 minutes, names your level and the one move that takes you up a rung. Free.
- 5
The dark software factory
“a black box that turns specs into software”
WritesAI
ReadsNobody
- 4
The engineering team
“write a spec” … “leave for 12 hours, and check to see if the tests pass”
WritesAI
ReadsTests, mostly
Level 3 to 4 guides - 3
The developer
“Your life is diffs.” … “almost everyone tops out here”
WritesAI, most of it
ReadsYou, full-time
The valley, levels 2 and 3 - 2
The junior developer
“where 90% of ‘AI-native’ developers are living right now”
WritesYou and the AI, paired
ReadsYou, every line
- 1
The coding intern
“you offload specific, discrete tasks to your AI intern”
WritesYou, mostly
ReadsYou, all of it
- 0
Spicy autocomplete
“not a character hits the disk without your approval”
WritesYou
ReadsYou
Start with the job in front of you.
Pick the situation you are in today. Each one has a place to start inside the toolkit.
-
WhenThe agent’s diffs are bigger than your attention span.
You get The Level 3 to 4 patterns: test gates, Stop hooks, sandboxes and review agents, worked out for Claude Code, Codex and Cursor, so a task ends with evidence instead of a wall of code.
-
WhenYour agent config is a paste from someone else’s repo.
You get The Setup Pack. Answer 9 questions about your stack and download
AGENTS.md,CLAUDE.md, Cursor rules and MCP config from 31 checked servers, plus an install prompt your agent runs to add the Stop hook and check every setting against what you have installed. -
WhenThe task is too big to hand over in one prompt.
You get The brief chain:
intent.md,spec.md,plan.md. The agent works from a spec you wrote and approved (spec-driven development), in the same order in Claude Code, Codex and Cursor. -
WhenThree tools ship changes every week and you can’t keep up.
You get Guides that cover Claude Code, Codex and Cursor side by side, a last-updated date on every page, and What’s New, a feed of what changed and which guides changed with it. What’s New is free for everyone.
-
WhenYour team works at five different levels.
You get The Team Toolkit covers up to 10 people for 12 months, for $299.99 once: you and up to 9 more, added by email address on your account page. Tech leads also get a reading path and the free Tech Lead Scorecard to place the whole team on the ladder.
-
WhenThe agent answered with a flag that doesn’t exist.
You get Codex has no
--approval-modeflag. The real one is--ask-for-approval, with the valueson-requestandnever(Codex CLI 0.159.2, checked 2026-10-02). The command reference lists stale flags the same way, each with the release it was checked against.
Everything in the Toolkit.
One payment covers the library, the Setup Pack and the books for 12 months.
The seven books, at a comparable ebook’s price$335
The Toolkit, paid once$59.99
-
The guide library
400+ guides and 100+ recipes for Claude Code, Codex and Cursor, filed by level, in English and Polish. They show how to write the
CLAUDE.mdandAGENTS.mdfiles your agent reads, one part of context engineering. Every page shows when it was last updated.Includedin the Toolkit
-
The Setup Pack
Nine questions about your stack, about 20 minutes, one zip generated for your repository. It sets up the harness around your agent, the subject of harness engineering.
- AGENTS.md
- CLAUDE.md
- .cursor/rules
- .mcp.json
- TESTING.md
- INSTALL-PROMPT.md
Includedin the Toolkit
-
Seven books
The four-title library, 328 chapters on 3,057 pages, plus three reading paths for developers, tech leads and CTOs. EPUB and PDF, English and Polish. Downloads are yours to keep. See the shelf
$335seven books
-
Free for everyone, without buying: 5 full articles a month (10 with a free account), the Start here and ladder guides, the three scorecards, the MCP server that lets your agent search the library, What’s New, the Discord for questions, and the public GitHub repo where you report a mistake.
Seven books, included in the Toolkit.
The four library titles run to 3,057 pages, and the three role paths reuse those pages in the order your role needs. At $47.99 each, the typical price of a comparable ebook, all seven would cost $335.
The AI Engineering Handbook
AI Engineering with Cursor
AI Engineering with Claude Code
AI Engineering with Codex
Agentic Engineering for Developers
Agentic Engineering for Tech Leads
Agentic Engineering for CTOs and Executives
1 / 7
Your first week, planned.
-
Day 1
Generate your Setup Pack.
Answer nine questions and download the zip. Your agent runs
INSTALL-PROMPT.mdand checks each setting against the version you have installed. About 20 minutes. -
Days 2 to 5
Hand over one real ticket.
Write
intent.mdandspec.mdfor something you were going to build anyway. Let the agent plan, build and stop on the tests. Read the report, then the code if you want to. -
By day 30
Decide with evidence.
You have one task done the new way. If that did not change how you work, email support within 30 days of buying for a full refund.
Buy the Toolkit Buy the Toolkit
$59.99 once for 12 months of access. Nothing renews on its own. 30-day refund on a first purchase.
Who checks this, and how.
# what a model suggested codex --approval-mode never # what Codex CLI 0.159.2 accepts codex --ask-for-approval never
-
Every page carries a date.
Each guide shows when it was last updated, so you always know how old what you are reading is.
-
The command reference names the releases it was checked on.
It states the Claude Code and Codex versions behind it, and its table of stale flags gives the release each row was checked against.
-
The Setup Pack catalogue is verified against its sources.
The 31 MCP servers in its catalogue were last verified against their upstream sources on 2026-10-02, 31 of 31. A weekly CI run is set up to repeat the check.
-
Numbers come from third parties, with a date.
Every statistic on this page names its publisher and when it was published. No multipliers that nobody measured.
-
Mistakes are reported in public.
Every guide has an Improve this page button that opens an issue in the public GitHub repo. When the author confirms it, the guide is corrected and its date changes.
Pay once, for you or for your whole team.
The Toolkit gives you 12 months of access, and nothing renews on its own. Prices in USD. The total for your country, with any VAT or sales tax, is shown at checkout before you pay.
Toolkit
Start hereFor a developer who wants a working agent setup and every guide for the climb.
$59.99once
12 months. Renew for $29.99 if you want more.
Buy the Toolkit Open your accountPaid once. 30-day refund on a first purchase.
- The full Setup Pack:
CLAUDE.md,AGENTS.md, Cursor rules,.mcp.jsonandTESTING.mdgenerated for your stack - 400+ guides and 100+ recipes for Claude Code, Codex and Cursor, each with a last-updated date
- All seven books, EPUB and PDF, English and Polish ($335 at the median price of comparable ebooks)
- Every update to the guides and the Setup Pack generator for 12 months
Team Toolkit
The same Toolkit for up to 10 people, on one company invoice.
$299.99once
12 months for the team. Renew for $149.99.
Buy the Team ToolkitPaid once. 30-day refund on a first purchase.
- Everything in the Toolkit, for each person on the team
- Up to 10 people: you and up to 9 more, added by email address on your account page
- One company invoice with VAT, issued by Polar
- Costs less than separate Toolkits from the sixth person
Services for teams
Fixed-price work on your repository, delivered by the author.
From $499
excl. VAT, invoiced by Wondel.ai
See services for teamsA free call first
- A repository diagnostic, with written findings and a next step
- A half-day workshop, or a working day with your team on one repository
- An audit and setup, or a pilot, with written acceptance criteria
-
Paid once
$59.99 for 12 months. Nothing renews on its own.
-
30-day refund
On a first purchase, for any reason. Email support and it is refunded.
-
Files you keep
Books and Setup Pack files you download stay yours after the 12 months.
-
VAT invoice
Polar, the merchant of record, sends it right after payment.
Who it is for.
A good fit if
- You use Claude Code, Codex or Cursor at work every week.
- You want to hand over bigger tasks and still sign off on what ships.
- You lead a team and want everyone working at the same level.
Probably not a fit if
- You want a promise that you’ll ship several times faster. The studies we cite show output and incidents rising together, so we make no such promise.
- You haven’t tried a coding agent yet. Start with the free ladder guide.
- Your team wants a plugin and won’t change how it works.
Agentic engineering vs vibe coding: every name, defined
Each of these names for building software with AI, and the one that means a different job, has its own page, defined and set against vibe coding.
Vibe coding vs agentic engineering, the long guide All glossary terms
Questions people ask before starting.
Something else? Write to support@developertoolkit.ai and the author answers.
Can’t I just ask the AI how to set this up?
You can, and the answer will often sound right. Models invent flags: ask for Codex approval settings and you may get --approval-mode, which does not exist. The toolkit keeps a command reference that names the release each stale flag was checked against, and every guide shows when it was last updated.
How is this different from the official docs?
Official docs describe one tool’s features. The toolkit describes the workflow across all three, filed by level: what to set up, in what order, and how to check the result.
How is agentic engineering different from vibe coding?
Karpathy published a transcript of his Sequoia Ascent talk (opens in a new tab) (30 April 2026), and one line of it sums up the difference: “Vibe coding raises the floor. Agentic engineering is about extrapolating the ceiling.” In vibe coding nobody reads the diff and no check stands in for the reader. The toolkit teaches agentic engineering: you review the spec and the test report, and checks you set up review the diff.
Is this AI-assisted coding or AI pair programming?
It starts there and goes further. AI pair programming is the way of working where you read every line the AI writes; AI-assisted coding is the umbrella over all of it, vibe coding included. The guides begin at pairing and move one loop at a time up the autonomy ladder, until checks you set up read the diff for you.
I only use one of the three tools.
That works. Each guide says which tool it covers, and the workflow is the same order in all three, so moving to another tool later costs you little.
Is it a subscription?
No. You pay $59.99 once for 12 months of access, and you are charged again only if you choose to renew.
What happens after 12 months?
Your access ends unless you renew. A renewal costs $29.99 and adds 12 months to your end date, or counts them from today if your access has ended, so renewing early costs you no days. The books and Setup Pack files you downloaded are yours to keep.
Do you offer refunds?
Yes, 30 days from your first purchase of the Toolkit or the Team Toolkit, for any reason. Email support@developertoolkit.ai and it is refunded. Renewals are not covered, as set out in the Terms.
Do I get an invoice for my company?
Yes. Polar, the merchant of record, issues an invoice with the VAT or sales tax for your country. It arrives right after payment.
Will agents replace me?
Your work moves to the spec, the architecture and the stop conditions. In the published transcript of his Sequoia Ascent talk, Karpathy says you are “in charge of taste, engineering, design, and whether the system makes sense.” The toolkit is about doing those well.
Hand your next ticket to a Level 4 setup.
“level 2, and every level after it, feels like you are done. But you are not done”
Dan Shapiro, “The Five Levels” (opens in a new tab), 2026-01-23
$59.99 once for 12 months of access. Nothing renews on its own. 30-day refund on a first purchase.