Last verified
AgentsPriced from the vendor's pageWe sell none of them

AI agents4

Coding agents, taken one at a time and finished — what every route in costs, what we could get the thing to do on our own machine, and the cases where the honest answer is to use something else. Where we have not run a tool ourselves the page says so instead of implying otherwise. The rankings above us in this search result are written by people selling one of the entries.

Tools covered
4
in depth, not listed
Capability runs
27
published with proof
Head-to-heads
5
on the tools we cover
Access routes
27
priced separately
In the queue
3
named, in order
Verified
08/19
2026
§ 01 / COVERED

What we've finished.

All capability runs →
Claude Code
Anthropic

Anthropic's coding agent · terminal, IDE, desktop and browser · one engine

Entry
$20/mo
Free tier
No
Ways to buy
8
Our runs
27
Verified 2026-08-19
Codex
OpenAI

OpenAI's coding agent · CLI, IDE, desktop, web and cloud · one ChatGPT account

Entry
Free, then $8/mo
Free tier
Yes
Ways to buy
7
Our runs
0
Verified 2026-08-19
OpenClaw
OpenClaw Foundation

Free software with a real bill attached — and the widest blast radius of anything we cover.

Entry
Free — you pay the model
Free tier
Yes
Ways to buy
5
Our runs
0
Verified 2026-08-21
opencode
Anomaly

Free software with fifty-two ways to attach a model — and a vendor whose own price claim we could check.

Entry
Free — you pay the model
Free tier
Yes
Ways to buy
7
Our runs
0
Verified 2026-08-28
§ 02 / METHOD

What "covered" means.

Full methodology →

Prices come off the vendor's page

Every figure is read from the seller's own pricing page and stamped with the date it was read. Nothing is remembered, converted, or averaged, and a tool sold four ways gets four prices rather than one convenient number.

Capabilities are run, not assumed

Where we can test a claim on our own machine we do, and the page carries the transcript. Where we cannot, the page says it read the documentation instead. Those two are labelled differently on purpose.

Nothing is scored

We publish no composite rating, no stars and no "8.4 out of 10". We run no benchmark that would justify one, and a number invented to fill a layout is worse than no number at all.

We sell none of them

We are not a vendor of any tool listed here, and no placement is for sale. That is not a virtue, it is just the reason our order can differ from the order on the pages ranking above us.

§ 03 / QUEUE

What's next.

In order of search demand, not preference. Each one gets the same treatment as the first — every route priced, the capabilities we can verify actually run — which is why they arrive one at a time rather than as a list of forty on day one.

  1. 01 Cursor already covered in six comparisons
  2. 02 Manus general agent, not a coding one
  3. 03 Cline open source, lives in the editor
§ 04 / ELSEWHERE

Other ways in.

One weekly digest. Zero noise.

One weekly digest · No spam, ever · Unsubscribe in one click