---
title: "Technical content marketing for developer tools"
description: "Technical content marketing agency for developer tools and AI products. Opinion, explainers, tutorials, and guides for the humans and machines that ship code."
url: "https://argot.dev/"
---

# Writing for developers. And their agents.

Opinion, explainers, tutorials, and guides, written for the humans and machines that ship code.

[Tell us what you're shipping](https://argot.dev/work-with-us#contact) [Read the work](https://argot.dev/portfolio)

Trusted by teams at

- ![Clerk logo](https://cdn.ploy.ai/816aa366-89ca-43d6-ab25-15835b756e06/user/e2e739d2-clerk.svg)
- Deno
- Modal
- ![Momentic logo](https://cdn.ploy.ai/816aa366-89ca-43d6-ab25-15835b756e06/user/f268de8e-momentic.svg)
- ![Neon logo](https://cdn.ploy.ai/816aa366-89ca-43d6-ab25-15835b756e06/user/0ff1e5f4-neon.svg)
- ![Retool logo](https://cdn.ploy.ai/816aa366-89ca-43d6-ab25-15835b756e06/user/4c946a69-retool.svg)
- Stream
- ![Hex logo](https://cdn.ploy.ai/816aa366-89ca-43d6-ab25-15835b756e06/user/9c88802e-hex.svg)

agent-pay.js

```js
import { wrapFetchWithPayment, x402Client } from "@x402/fetch";
import { ExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";

const agent = privateKeyToAccount(process.env.AGENT_KEY);
const payees = new Set([process.env.QUOTES_WALLET.toLowerCase()]);

const client = new x402Client()
  .register("eip155:8453", new ExactEvmScheme(agent))
  .setSpendControls({ maxAmountPerPayment: "$0.05" })
  .onBeforePaymentCreation(async ({ selectedRequirements }) => {
    const payTo = selectedRequirements.payTo.toLowerCase();
    if (!payees.has(payTo))
      return { abort: true, reason: `Unknown payee ${payTo}` };
  });

const payFetch = wrapFetchWithPayment(fetch, client);
const res = await payFetch("https://api.example.com/v1/quotes");
```

1. **Give the agent its own wallet** (line 5). Every payment gets signed with this key, so don’t give the agent a wallet with anything else in it. Set up a new one and fund it with only what you’re happy for the agent to spend. It doesn’t need any ETH either. What the agent signs is an authorization to move USDC, and a facilitator submits that on chain and covers the gas.
2. **Name the network by CAIP-2 ID** (line 9). eip155:8453 is Base mainnet. x402 v2 names networks by CAIP-2 ID, where v1 used names like base-sepolia. If you copy a v1 snippet, the client ends up registered for a network name the server never uses, and the 402 that comes back won’t match it.
3. **Cap each payment** (line 10). Out of the box the client won’t sign anything over $1\. That’s a lot for one API call, and an agent stuck in a loop will make the same call over and over. Set the cap to what a single call should cost. This only limits each payment, though. If you want a daily limit, you have to track that yourself.
4. **Check who you're paying** (lines 11 to 15). The 402 response tells the client which wallet to pay, and a spoofed or compromised endpoint can put any address in there. Check that payTo is one you’ve approved and abort if it isn’t. Lowercase both sides first, since EVM addresses often turn up in mixed-case checksum form.
5. **Then it's just fetch** (lines 17 to 18). From here the agent calls the endpoint like any other. The first attempt comes back as a 402 with a PAYMENT-REQUIRED header. The wrapper signs the payment, retries with a PAYMENT-SIGNATURE header, and gets the data back with a receipt in PAYMENT-RESPONSE. It won’t pay twice for one request.

> “Can't recommend Argot enough. They have deep technical knowledge—you get authentic, developer-to-developer material that adds value. They can write thought-leadership just as well as technical content, which is very hard to find. Responsive and never miss a deadline.”

Carlota Soto, PMM, Neon (now Databricks)

## Thought for developers

Developers are working at a higher level of abstraction. They need the why and the what more than the how.

[How we work](https://argot.dev/work-with-us)

### Opinion

What you think. Why you built it, what everyone else gets wrong, and where this goes next. Developers choose the product with a point of view.

### Explainers

What you know. Intuition from your years on the problem, set out so developers learn from the subject matter experts.

### Tutorials

What you do. A real build, start to finish, so developers see what’s possible before they decide what to build.

### How-to guides

How it works. One problem and how your product fixes it. Enough to trust it with their problems.

## Run by agents

Agents are doing the how now. They need it exact, short, and working the first time.

### how-tos.md

How it works. An agent hits one problem mid-task and needs it fixed now. The exact fix, runnable as written.

### tutorials.md

What you do. An agent using your product for the first time has only your tutorial to go on. Every step runs in order.

### explainers.md

What you know. An agent choosing an approach needs the concept in its context.

### opinion.md

What you think. When a developer asks their agent what to use, it answers with what it has read.

## Blank page, rough draft, or no plan yet?

We write from scratch, fix what you have, and work out what to publish next.

Writing

We write with a terminal open.

A tutorial that doesn’t run is worse than no tutorial. An essay by someone who’s never used the product isn’t far behind. We install it, build something with it, and write from there, how-tos and essays both.

Editing

Cut the slop from your drafts.

You should use Claude to write. You shouldn’t publish it straight from the LLM’s mouth. We edit both strategically (is this an accurate thesis?) and tactically (is this an AI-ism?) to make it worth reading.

Strategy

SEO. AEO. GEO. LEO. Wait, that’s satellites.

To be used, you must be found. Whether that’s by a human, a chatbot, or a coding agent. We work from your briefs or prepare our own based on the best keywords, questions, and prompts to get you ranked, cited, and picked.

## Written to be found, read, & run

Five rules every piece follows, from the first outline to the final edit.

### Why before how

// Agents can handle the how. Your content has to answer why.

### Written for two readers

// A developer reads it to decide. An agent reads it to act. If either can’t follow it, it isn’t ready.

### If it's written, it runs

// Every snippet and command on the page has run in a terminal.

### Models draft, people finish

// LLMs are great for getting ideas down, not out. Taste is still human.

### Clear beats clever

// Anyone (and anything) can be clever; writing has to be clear to be useful.

## Notes on technical writing

Mostly about writing. Increasingly about agents.

[Read the blog](https://argot.dev/blog)

[Machine-to-Machine Marketing](https://argot.dev/blog/machine-to-machine-marketing)

[A minimum viable content strategy](https://argot.dev/blog/minimum-content-strategy)

[What is technical content marketing?](https://argot.dev/blog/what-is-technical-content-marketing)

## Developers who can write, we pay you to

Who pays technical writers? We do. Send a few links to technical writing you are proud of, with bonus points if you built the thing you wrote about.

[Apply to write](https://argot.dev/work-for-us)

## Questions

Short answers about what we write and how we work.

### What is Argot?

Argot is a technical content marketing agency for developer tools and AI products. It writes, edits and plans tutorials, how-to guides, explainers and opinion for developers and the agents they use.

### What does Argot write?

Tutorials, how-to guides, explainers, comparison pages and opinion, including essays under your founder's name. Argot also edits drafts your team or your model wrote, refreshes old tutorials, and plans what to publish.

[See the published work](https://argot.dev/portfolio)

### What does writing for developers and their agents mean?

Developers now hand docs and tutorials to coding agents that read them and act on them. Argot writes each piece for both readers: the why comes first for the developer, and every step and command is tested so an agent can follow it as written.

### How does Argot make sure the code in its content works?

Argot builds with the client's product before writing, and runs every line of code in a draft before the client sees it.

### Which companies has Argot written for?

Argot has written for devtool and AI companies including Clerk, Deno, Modal, Momentic, Neon, Retool, Stream and Hex.

[Read the published work](https://argot.dev/portfolio)

### Why use Argot instead of freelance writers?

Hiring writers one by one means testing them, explaining your product, and fixing drafts that miss the technical details. Argot learns your product and tests every example, so the first draft already holds up to technical scrutiny.

[Compare agencies, freelancers and in-house writers](https://argot.dev/agency-vs-freelancers-vs-in-house)

### What does Argot cost?

Every engagement is scoped to your budget. Tell Argot roughly what you have to spend and what you need, and it proposes the mix of pieces and services that fits.

### How do I start working with Argot?

Fill in the short form on the Work with us page or email work@argot.dev. Argot comes back with ideas for what to write, then talks through your product, your readers and your goals before writing.

[Start a project](https://argot.dev/work-with-us#contact)

## What should developers build with your product?

Send us a few lines about the product, who uses it, and what you want them to build. We'll come back with ideas for what to write.

[Tell us what you're shipping](https://argot.dev/work-with-us#contact) [work@argot.dev](mailto:work@argot.dev)
