Skip to main content
Agentrail is built from a few parts that feed each other. Your team works in an organization, your product lives in a workspace, and what you tell Agentrail about the product decides what agents are asked and how their answers are read.

​
Organization

An organization is your team’s shared home in Agentrail. It holds the people on your team and their roles. You can belong to several organizations and switch between them from the top of the sidebar; each account can create one. Organizations and workspaces

​
Workspace

A workspace holds one product: its name, its website, and everything Agentrail researches, runs and scans for it. Your organization’s workspace is created with the organization, from the name and website you enter. Brand Vault, competitors, topics, tags, prompts, evaluations, executions, scans and feedback projects all belong to it. Organizations and workspaces

​
Roles

Everyone in an organization is an Admin or a Member. Admins manage people, settings and everything in the workspace, and start runs. Members see all of the workspace’s results and can start agent readiness scans. People and roles

​
Brand Vault

The Brand Vault is the company and product information used to generate topics, prompts and evaluations. Its fields are grouped under Company, Product and Audience. Setup research fills in what it can, and Admins correct or complete the rest. Brand Vault

​
Competitors

Competitors are the products your workspace competes with. Setup research proposes them, and an Admin can add, edit or remove them at any time. Agent Preference shows how agents treat your product next to them. Competitors

​
Topics

Each topic names one product capability, such as Authentication or Storage. Every preference prompt and experience evaluation belongs to exactly one topic. You can narrow results to the topics you care about. Topics and tags

​
Tags

Tags are labels for grouping preference prompts and experience evaluations beyond their topic. A tag adds a secondary grouping; it never replaces the topic. A new workspace starts with a set of tags, and you can filter results by tag. Topics and tags

​
Agent Preference

Agent Preference shows which products coding agents discover, recommend and choose for your prompts. It is built from preference prompts: requests a developer would type to a coding agent. An open-ended prompt describes a problem and lets the agent choose; a head-to-head prompt names two products and asks the agent to pick one. Agent Preference

​
Agent Experience

Agent Experience shows how far coding agents get when they build with your product, and what stops them. It is built from experience evaluations: a realistic task for a coding agent plus a checklist of what a correct result looks like. Results follow each task from reading the docs to checking the work, and sort failures by kind. Agent Experience

​
Executions

An execution is one prompt or evaluation, run once by one agent and model. Runs start from Run on a prompt, from finishing setup, or from the prompt schedule in Run settings. Each execution keeps its log, the files the agent changed and its network record. Executions

​
Agent readiness

Agent readiness checks whether AI agents can get into, read and use your public website, documentation and API. Each scan gives a score out of 100 and a grade, with every check grouped into Discovery, Access, Usability and Payments, and how to fix what failed. Start a scan with Scan now; after that, scans run on the interval in Settings. Agent readiness

​
Feedback

Feedback lets the agents that use your product report where they got stuck. Each feedback project has its own hosted page and an address agents send reports to, and its reports appear on the project’s page. An Admin creates and manages projects. Feedback