What Is AgentRel?

TL;DR

• AgentRel describes how AI agents discover and use developer products on a person's behalf. Agent-Led Growth is the more widely accepted term for that broader growth motion.
• AgentRel is a newer name for the same motion.
• Agent discoverability maps to Agent Preference Optimization (APO), which covers whether agents find and choose your product. Agent usage or usability maps to Agent Experience (AX), which covers whether they can make it work.
• The guide explains how agent recommendations become working implementations and how you can measure each step.

What is AgentRel for developer tools?

AgentRel describes how AI agents discover developer tools and use them on someone's behalf. A coding agent can find an API, choose its SDK, and try to integrate it into a project. Agent-Led Growth, or ALG, is the more widely accepted name for this growth motion. AgentRel is a newer name for it, not a separate process.

One use of the term describes AgentRel through "agent discoverability" and "agent usability." Gauge uses Agent Preference Optimization, or APO, for work that helps agents find and choose a product. It uses Agent Experience, or AX, for work that helps agents install and use that product successfully. People also use "agent usage" or "agent usability" for this part of the work. These labels do not establish an industry-wide standard.

How does AgentRel turn recommendations into successful implementations?

A coding agent can recommend your tool without choosing it for the task. In Gauge's Agent-Led Growth framework, recommendation, selection, installation, and successful implementation are separate outcomes. For example, an agent might name your API as an option but select another provider because the repository already uses it.

Installation still falls short of a working integration. A package may appear in the project while authentication fails or the agent leaves code that cannot complete the requested task. If you count every mention as adoption, you cannot tell whether the agent passed over your tool, stopped during setup, or finished the implementation.

Agent Preference Optimization, or APO, covers the earlier decisions. It asks whether agents recommend and select your product for relevant tasks. Agent Experience, or AX, covers what happens once an agent tries to use it. To find where progress stops, check the agent's choice, package installs, and final code separately rather than rolling them into one success metric.

How does AgentRel discoverability affect coding-agent tool selection?

A coding agent can draw on model knowledge, the current task's context, and live research when choosing a tool, as Gauge's Agent-Led Growth guide explains. Its model has knowledge from training, while the prompt, repository, and rules files provide context for the task. The agent may also search the web or fetch documentation.

Those inputs operate on different time scales. Changing a page an agent can find may affect its live research, but it does not change the model's prior knowledge or a repository rule that specifies a provider. You need to know which input shaped the choice before deciding what to fix.

A fetched page also needs to help the agent use the product. If an agent finds your SDK but the documentation leaves out setup or configuration steps, the agent may select it without producing working code. Search visibility can put a product in front of an agent, but the documentation it retrieves must also give it the steps needed to proceed.

How does AgentRel differ from agent-led growth, SEO, AEO, and GEO?

AgentRel and Agent-Led Growth describe the same broad motion for developer tools. Agent-Led Growth is the more widely accepted name for growth driven by agents that discover, select, and use a product. AgentRel is a newer name for the same motion.

SEO helps your pages appear in search results. AEO and GEO focus on how AI-generated answers represent and cite your product. Those practices can help an agent discover your tool, but they do not show whether a coding agent selects it or produces working code with it.

AgentRel covers both discovery and use. Agent Preference Optimization (APO) addresses whether agents find and choose your product. Agent Experience (AX) addresses whether they can implement it. If an agent recommends your SDK but fails during setup, better answer visibility will not fix the setup failure.

What the data shows about agent research behavior

In Gauge's analysis of 500 coding-agent runs, documentation accounted for 55% of fetched sources. Source code accounted for 18%, package registries for 11%, and third-party content for 5%. These percentages describe fetched sources in that sample, not the share of agents that used each source or completed an integration.

The Agent-Led Growth guide also reports fetched setup pages, READMEs, quickstarts, and API references. Check the source's category definitions before comparing those page types with the broader source breakdown. Package registries appear in both descriptions, so the figures should not be added together as one split.

For agent experience, start by checking setup, README, and quickstart pages when an implementation stalls. Agents fetched those page types frequently in this sample, but a fetch count cannot tell you which page caused a failure.

How do you measure and improve AgentRel discoverability and usability?

Start with two baselines. For discoverability, give coding agents open-ended tasks across prompts and repositories that reflect your customers' work. Record recommendations and selections separately, then track installations as the next step toward implementation. For usability, give agents tasks that name your product and check whether they leave working code. Track fetched pages and the share of recommendations that become installations, since a recommendation alone does not show that an agent can build with your tool.

Inspect failed runs before changing anything. Review search and fetch traces alongside the agent's choices and execution logs. Together, they can show whether it missed a relevant page, chose a competitor, followed outdated instructions, or stopped at an API error. Fix the specific page, instruction, or product behavior that blocked the task. The Agent-Led Growth guide describes this measurement loop in more detail.

Rerun the same task with the prompt, repository, agent configuration, and pass criteria held constant. Repeat the test across more runs and configurations before treating an improvement as general. Gauge Agents connects discovery decisions to coding-agent implementation through isolated sandboxes, session traces, and controlled reruns. Compare any improvement in test results with signups, activation, and retained usage if you can track those outcomes. A successful agent run shows that an implementation worked under the tested conditions, not that it became an active customer.

The takeaway

Measure whether coding agents choose your developer tool and whether they finish a task with it. Gauge Agents helps you inspect where a run stopped and repeat the task after a fix. For the broader framework, read Gauge's Agent-Led Growth guide.

FAQ

Is AgentRel the same as Agent-Led Growth?

Both terms describe growth driven by agents that discover and use products on a person's behalf. Agent-Led Growth is the more widely accepted name for the broader motion, while AgentRel is a newer alternative. The Agent-Led Growth guide explains the framework in more detail.

What should a developer-tool company prioritize first in AgentRel?

Start with open-ended coding tasks to see whether agents choose your product without being told to. Then name your product in an integration task and check whether the agent leaves working code, even if it did not choose the product on its own. Gauge Agents records the decisions and implementation steps so you can investigate where runs stop.

How is AgentRel different from traditional developer marketing?

Traditional developer marketing usually helps a person find and evaluate a tool. AgentRel also considers what happens when a coding agent chooses a tool and tries to implement it. An agent might recommend your API but fail to install its SDK or produce working code.

How do I measure AgentRel discoverability for my developer tool?

Run open-ended tasks that do not name your product, and count how often agents recommend and select it against alternatives. Compare results across the coding agents and repositories your customers use. Gauge Agents records each run so you can inspect which product the agent chose and what information it used.

Where does Gauge fit in?

You can measure agent discoverability by checking whether coding agents recommend and select your product. You can measure usability by checking whether they complete a working implementation. Gauge Agents runs coding agents on real tasks and records their steps so you can diagnose failures and test changes.

What does an AgentRel usability test for an API integration involve?

Give a coding agent a defined integration task with a clear pass condition, then check its installed package and final code. Review where it needed human input or failed to recover from an error. Gauge Agents runs these tasks in isolated workspaces and records the steps behind each result.