# Agent registries

URL: /workflows/plays/agent-registries · Id: `play:agent-registries`

> Agent registries list every Startup’s tools where Agents and developers search, with tool descriptions tested against agent tool search.

Agent registries list a [Startup](/business/nouns/startup)’s tools where [Agents](/business/nouns/agent) and developers search, such as the MCP Registry. For an agent buyer, the tool description is the whole storefront.

## Ratings
| Dimension | Rating | Reads |
| --- | --- | --- |
| Scalability | 5 of 5 | very high: higher grows further with spend or automation |
| Cost | 1 of 5 | very low: lower costs less per Customer |
| Latency | 2 of 5 | low: lower answers sooner |

## Budget
| Property | Value |
| --- | --- |
| Tier | $0 |
| Minimum effective budget | $0 a month |

## How it runs
Every Live Startup with an MCP or API [Interface](/business/nouns/interface) is listed by default. A tool description ships only when test agents pick the right tool for the right task.

## Processes
Runs as [Launches](/workflows/processes/launches), accountable to the [Marketing Team](/agents/marketing).

## Fits
| Fits | Value |
| --- | --- |
| Family | demand |
| Shelf | Seed: finds the first Customers (Launch) |
| Moves | [Organic share](/agents/marketing#organic-share) |
| Motions | [b2a](/business/motions/b2a), [b2a2b](/business/motions/b2a2b), [b2a2c](/business/motions/b2a2c), [b2a2d](/business/motions/b2a2d), [b2d](/business/motions/b2d) |
| Venues | [MCP Registry](/workflows/plays#mcp-registry), [ChatGPT](/workflows/plays#chatgpt), [Claude](/workflows/plays#claude) |
| States | [Live](/business/lifecycle/states/live), [Repeatable](/business/lifecycle/states/repeatable), [Scale](/business/lifecycle/states/scale) |

## Loop
| Step | Value |
| --- | --- |
| Hypothesis | Tool descriptions tested against agent tool search bring more organic first touches from Agents than untested descriptions. |
| Measure | [Organic share](/agents/marketing#organic-share) over 30d, at least 50 observations |
| Learn | The learn step reads the Readout, records Evidence and a Decision, and drafts the next Variant. |

## Example
A Startup selling address validation lists one tool, `validate_address`. Its description names inputs, coverage and price per call, and test agents choose it over three rivals on 46 of 50 tasks.