# A developer marketing agency vs agent discoverability

> Draft.dev, Hackmamba and Infrasity produce good technical content. None can tell you whether Claude Code installed your package.

Source: https://armature.tech/library/developer-marketing-agency-vs-agent-discoverability
Published: 2026-09-03
Publisher: Armature, Inc. (https://armature.tech)

---

The good developer content agencies are genuinely good. Draft.dev writes with engineers rather than copywriters and does it at volume. Hackmamba and Infrasity are real operations with real output. If you need forty solid technical posts, hire one of them.

This page is about what that does and does not buy you now.

## The short answer

> A **developer marketing agency** produces the content. **Agent discoverability** measures whether the content changed what a coding agent installs, and works the levers content cannot reach. The first is production. The second is measurement plus a different set of levers.

## What content agencies do well

- **Volume with technical credibility.** Engineers writing for engineers, at a cadence an in-house team cannot sustain.
- **Documentation and tutorials.** The highest value pages for a developer tool, and the ones most companies neglect.
- **Search fundamentals.** They know how to structure a page and target a query.
- **They ship.** The main reason to hire an agency is that the content actually appears.

None of that is obsolete. Content is still a lever.

## What changed underneath it

The content agency model assumes a person finds the content, reads it, and decides.

A growing share of the decisions now happen without that. A developer opens a repository in a coding agent, types one line, and the agent reads the codebase, works out what is needed, sometimes searches, installs a product and writes the integration.

Two consequences.

**The content might not be read at all.** On decision tasks, Claude Code ran a web search in 1.6% of runs. Codex ran one in 53%. Against one of those, forty excellent posts are absent from the moment of choice.

**The largest lever is not content.** In our tests, the repository decided more than anything published. The same hosted database request produced one winner in 111 of 111 JavaScript sessions and 48 of 132 TypeScript sessions. And a written instruction naming a preferred product lost 2 times out of 3 once a rival library was installed in the same repository.

No writing pipeline reaches either of those.

## The four levers, and who works them

| Lever | Size | Content agency | Agent discoverability |
| --- | --- | --- | --- |
| What is in the repository | Largest | No | Yes: templates, framework integrations, official examples, platform defaults |
| Which agent is running | Very large | No | Measured, and the spend matched to it |
| Who is asking | Large | Partly, if told which persona | Measured per persona |
| What the agent can read | Moderate | Yes, this is the whole offering | Yes, and directed by what the losses show |

A content agency owns the fourth row well and does not touch the first three.

## Side by side

| | Developer marketing agency | Agent discoverability |
| --- | --- | --- |
| Output | Blog posts, tutorials, docs, comparison pages | A measured install share, the changes that move it, and the changes shipped |
| Measurement | Traffic, rankings, sometimes signups | Install share by agent, language and persona |
| Feedback loop | Analytics, weeks to months | The agent's own recorded reason for rejecting you |
| Levers beyond content | Few | Templates, integrations, package naming, quickstart correctness |
| Typical price | Per article, or a monthly retainer | From $5,000 per month, with a growth engineer |

## The feedback loop is the real difference

This is the part worth paying attention to.

When a content agency publishes forty posts, you find out whether they worked from traffic and, with luck, attribution. That signal is slow, noisy and indirect.

When you run real agents, the losses come back with reasons. A judge records, for every session, which products were raised, which were rejected, and why. In our data those reasons cluster into a short list:

- The quickstart did not run, so the agent abandoned the integration part way through.
- The documentation described an interface that had changed.
- The package name did not match the product name and the install went wrong.
- There was no answer to a specific production question the agent asked.
- The fit was genuinely wrong for that repository.

Four of those five are fixable in days, and none of them are fixed by writing more posts. They are fixed by editing three pages.

That is why the order matters: measure, then commission the content the losses point at.

## The honest recommendation

If you have no content, hire an agency. An empty documentation site loses on both surfaces and nothing else you do will compensate.

If you have content and do not know whether it works, measure before you commission more. Run 300 agent sessions yourself: three repositories that look like your users' projects, ten requests written as symptoms, five runs each with Claude Code and Codex. A weekend and a few hundred dollars in tokens.

Then read the losses. In our experience the list of fixes that comes back is shorter and cheaper than the content brief you were about to write.

## Common questions

### What does a developer marketing agency do?

Produces technical content at volume for developer tool companies: blog posts, tutorials, documentation, comparison pages, usually written by engineers rather than copywriters. Draft.dev, Hackmamba and Infrasity are established examples.

### How is agent discoverability different from hiring a content agency?

A content agency produces the artefacts. Agent discoverability measures whether those artefacts change what a coding agent installs, and it works the levers that content cannot reach, starting with what is already in the customer's repository.

### Do I need both?

Often yes. Content is a real lever and volume helps, especially against agents that search the web. The measurement is what tells you which content to produce and whether any of it moved the number.

### Can a content agency measure install share?

Not without running real coding agents on real tasks at scale, which is a different kind of operation from a writing pipeline. Each session is minutes of agent time and hundreds are needed per category.

## Read next

- [Developer tool marketing in the agent era](https://armature.tech/library/developer-tool-marketing-in-the-agent-era) (Markdown: https://armature.tech/library/developer-tool-marketing-in-the-agent-era.md)
- [Agent discoverability: the complete guide](https://armature.tech/library/agent-discoverability) (Markdown: https://armature.tech/library/agent-discoverability.md)
- [Technical content marketing when the reader is a machine](https://armature.tech/library/technical-content-marketing-for-ai-search) (Markdown: https://armature.tech/library/technical-content-marketing-for-ai-search.md)
- [The best AI visibility tools for developer tools in 2026](https://armature.tech/library/best-ai-visibility-tools-for-developer-tools) (Markdown: https://armature.tech/library/best-ai-visibility-tools-for-developer-tools.md)

---

Armature helps software products get discovered and used by coding agents.
Service: https://armature.tech/discoverability · Results: https://armature.tech/leaderboards/sectors · Contact: contact@armature.tech