The short list of AI agent tools that actually work.

Agentic coding and ops in the terminal.

Runs my daily operations: reports, deploys, file work, whole pipelines. The tool I measure every other agent against.

BEST FOR

real work in real repos, end to end

SKIP IF

you want a chat window, not a worker

# coding-agents

One protocol to wire agents into your tools.

Design files, analytics, email, hosting. One standard instead of custom glue per app. This is where agent stacks are heading.

BEST FOR

connecting agents to an existing stack

SKIP IF

you automate a single app

# mcp-servers

Browser control agents can rely on.

Logins, scraping, UI checks. Handles real-world pages and auth better than anything else I put into rotation.

BEST FOR

web automation behind a login

SKIP IF

plain HTTP gets the data

# browser-automation

The production stack behind my SEO agent.

Claude Code, Ahrefs, Search Console, Playwright and a custom agent. Each tool has one job, a human approves what ships.

BEST FOR

running SEO as an agent workflow

SKIP IF

you publish twice a year

# custom-agents

Paid social under full agent control.

Full scope: campaigns, ad sets, creatives, audiences, budgets and reporting. Agents run the ops, a human signs off on spend.

BEST FOR

automating Meta campaigns end to end

SKIP IF

you run one campaign and edit it by hand

# ad-platforms

Search campaigns without the console grind.

Full scope: campaigns, keywords, bids, budgets and reports. Same rule as Meta, agents operate, humans approve spend.

BEST FOR

search ops at scale

SKIP IF

spend is too small to automate

# ad-platforms

No-code glue that keeps running for months.

Lead forms to sheets, webhooks to CRMs. Scenarios I built once and stopped thinking about.

BEST FOR

wiring SaaS apps without code

SKIP IF

the logic fits in thirty lines of code

# orchestration

Visual workflows with an agent layer on top.

The AI nodes matured enough to carry production pipelines. Visual ops the whole team can read.

BEST FOR

teams that want visual ops

SKIP IF

your team prefers code-first

# orchestration

Real search queries and positions, from the source.

Every content decision on this site starts from it: real queries, real positions, and whether Google indexed the page at all. Estimates describe the market; this describes my pages.

BEST FOR

planning content from measured demand

SKIP IF

the site you care about is not yours

# data-scraping

Design files agents can read, not guess.

Entire landing pages went from mockup to deployed code through it. Node ids make prompts precise, tokens survive into the code, and the visual check stays human.

BEST FOR

design-to-code with a real design file

SKIP IF

the project has no design artifact

# mcp-servers

The channel my agents report through.

Background work is invisible until a message says what happened. Daily numbers to a group, failures to my DM, one POST request each, read within minutes.

BEST FOR

closing the loop on every automated job

SKIP IF

your team does not live in Telegram

# monitoring-evals

The table humans and agents co-author.

Agents append rows and fill blanks, people read, filter and correct. Append-only writes plus a dedup key made months of reruns boring, with zero data lost.

BEST FOR

reports non-technical people actually use

SKIP IF

no human ever opens the sheet

# data-scraping