How to get your Claude Connector organically discovered

Imagine you have a Claude Connector. You can build it, publish it and make it live in Claude's Connector Store.

Until now, being live did not mean customers would discover you. A customer generally had to open the Connector Store, browse or search for a solution, choose your Connector and connect it. If they did not already know your product existed, they were unlikely to find it.

That experience is changing. When someone asks Claude an action prompt such as book a hotel, find a job or update a project, Claude can present a shortlist of relevant Connectors at the moment they are needed, even when the customer has never heard of them. This is organic Connector discovery.

If your Connector appears, you can win a customer at the moment they are ready to act. If it does not, a competitor may be selected before the customer visits your website or compares your product.

This is a measurable and optimisable surface. By working with AgentDiscoverability.com, DirectBooker increased its pick rate from 31.5% to 100% across its highest-priority hotel-search prompts. Across the wider category, it moved from #5 to #2.

Agent Tool Optimisation

This is a different discovery layer and a different optimisation surface. Answer Engine Optimisation (AEO) can shape what an AI says about a company, but it does not determine which third-party tools an agent discovers, selects and ranks.

Agent Tool Optimisation (ATO) is the practice of making a business's capabilities discoverable, selectable and reliably invoked by agents acting on behalf of users.

Here is the four-stage approach we use to optimise for it.

First, build an agent-ready integration

Build and publish an agent-ready version of your product so customers can discover and use it inside Claude. That means turning your product's capabilities into clearly named tools, descriptions and schemas Claude can understand and call.

Once the Connector is live, you can measure whether Claude discovers it for the right customer requests.

1. Measure

Run the prompts your customers use across ChatGPT, Claude and other agent surfaces.

Prioritise the commercially valuable customer intents your business cares most about winning, then use representative category prompts to establish a reliable baseline.

Run each prompt repeatedly and track three stages:

• Found - did your Connector appear in Claude's connector search?
• Picked - did Claude place it in the connector picker?
• Positioned - where did it rank?

We combine these signals into a Discoverability Score out of 100. It shows how consistently Claude finds, picks and ranks your Connector for the customer intents that matter.

Behind the score, you can see the prompts you win, the prompts you miss and the competitors Claude presents instead.

2. Diagnose

Identify the stage where your Connector loses visibility and what needs to become clearer.

Each type of loss points to a different problem:

• Not Found - Claude is not consistently connecting the customer's intent with your Connector.
• Found but not Picked - your Connector appears, but its fit for the task is not clear enough to make the shortlist.
• Picked but poorly Positioned - your Connector is relevant, but competitors are being presented earlier or more consistently.

Look for patterns across repeated priority prompts rather than treating one result as the answer. Then compare the agent-facing descriptions, capabilities and supporting information of the Connectors that win more often. This shows whether the main opportunity is clearer relevance, clearer task fit or stronger consistency.

3. Improve

Make focused changes that address the specific stage where visibility is being lost.

Turn the diagnosis into a short, prioritised set of updates to the way your Connector explains what it does, which customer tasks it supports and when Claude should use it.

The right improvement may be clearer listing language, more legible tool names and descriptions, better supporting documentation or a capability gap that needs to be closed. Change the highest-impact area first rather than rewriting everything at once. Record exactly what changed and when it went live so its effect can be measured cleanly.

4. Verify

Run the same priority prompts again and prove that Found, Picked and Positioned outcomes improved.

Once the changes are live, rerun the original prompt set under comparable conditions and measure it against the baseline. The goal is to see repeated movement at the stage you set out to improve, without losing prompts you were already winning.

A single favourable result is not proof. Look for a consistent lift across repeated runs, priority prompts and the overall Discoverability Score. Keep monitoring after the initial improvement because models, Connector listings and competitor signals continue to change.

What happened with DirectBooker

DirectBooker worked with AgentDiscoverability.com to apply this process to its Claude Connector. Its pick rate across priority prompts increased from 31.5% to 100%, and it moved from #5 to #2 across the wider Hotel Search & Booking category.

DirectBooker used our AgentDiscoverability.com product range end to end.

• Measurement Engine showed where its Connector was found, selected and ranked across high-value hotel-search intents.
• Optimisation Intelligence identified and prioritised the changes most likely to improve those outcomes.
• Accelerator which is where our team worked directly with DirectBooker to implement the priority optimisations, measure the right signals and verify the uplift, always with the goal of making sure they rank top for their target prompts, consistently, over time.

DirectBooker now outranks some of the world's largest travel companies for organic Connector discovery. It shows that investing in effective Agent Tool Optimisation can create a genuine commercial advantage, and that Connector discovery can be measured and improved.