Skip to main content
ish is a simulated human experience engine. Simulated people experience the thing you are making (a link, an ad, a video, a PDF, a chatbot) and report back what they noticed, where they got stuck, and what they would do next, before it ships. The output is a reported journey: sentiment, friction, blockers, positive moments, completion, each carrying the reasoning behind it. You see why a reaction happened, not just a number. An ish of someone is close enough to make a better decision, not a precise prediction.
Reach for ish when the question is “will this land before we put it in front of people?” It is not analytics (that tells you what already happened), not a survey, not a scoring tool, and not a chatbot you talk to. Simulated people experience your artifact and report back.

​
Install, connect, run, share

Install

Drop the CLI on your PATH with one line, or wire ish into your agent over MCP. Install the CLI.

Connect

Point a client at the hosted MCP server, or sign in from the terminal with ish login. Connect an agent.

Run

Create a study, draw an audience, and simulate a visit. Run a study end to end.

Share

Hand anyone a public, no-login link to the reported journey. Share results.

​
The shape of a run

Every run follows the same arc, whichever surface you drive it from:
  1. A workspace holds your studies, people, and sources.
  2. A study is the persistent unit: a modality plus what you want people to react to.
  3. An iteration carries the artifact for one run (a URL, a file, or a chatbot endpoint).
  4. People are the simulated audience, generated from a brief and optional grounding sources.
  5. A run sends the people through the study and reports back a journey for each one.
A study covers any pre-ship artifact, not just web pages: interactive (a URL), text, video, audio, image, document, and chat (probe an external chatbot endpoint).

​
Choose how you build

One backend, two independent surfaces — the CLI and the MCP server. You don’t need both; pick the one that fits where you work. The CLI runs on your machine (local development, scripts, local coding agents); the MCP server is for agents, including hosted ones with no access to your machine.

The ish CLI

Install @ishlabs/cli, sign in, and simulate your first visit from the terminal. The local surface — built for a prompt and for CI, and the only way to test a web, iOS, or Android app running on your own machine.

ish over MCP

Connect any agent — including hosted assistants with no access to your machine — to the MCP server at mcp.ishlabs.io/mcp, then run studies as agent-native tools.

Your AI client

Claude, Cursor, ChatGPT, VS Code, and more. Each connects to the MCP server and signs in the same way.
Each surface works on its own, so you never need both. They do share one backend, so if you happen to use both, a study you create in one shows up in the other.

​
Connect an AI agent

The MCP server wires into the clients you already use. OAuth runs on first connect.

Claude

Cursor

ChatGPT

VS Code

Lovable

Replit

​
Where to go next

Quickstart

Install ish and read your first reaction in under five minutes.

CLI reference

Every command and flag, grouped by namespace.

MCP reference

Every tool and resource the hosted server exposes.

The ish API

Put a person in the decision loop of your own environment, turn by turn.

Get ish free

Create an account and keep your studies.