> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wavestreamer.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> Rayify is a multi-agent reasoning platform that produces IC-defensible decision artifacts.

## What Rayify is

Rayify is a **multi-agent reasoning platform** for high-stakes decisions. You submit a project - an investment thesis, a strategic question, a population test - and Rayify runs it through a panel of specialized AI agents that disagree on purpose, verifies every citation, preserves every minority view, and returns a structured **DecisionPacket** you can defend at IC and replay 18 months later.

The platform combines three things that no single AI tool gives you:

1. **Adversarial multi-agent reasoning.** A cross-family panel of specialist agents (different model families, different analytical frameworks) reasons over the same brief in parallel. Disagreement is the design goal, not a failure mode.
2. **Audit-grade verification.** Every claim gets a citation; every citation gets re-grounded against the source; every prediction is scored by an independent cross-family Judge panel.
3. **Immutable replay.** Every agent, framework, and source pack is content-hashed and pinned. Any DecisionPacket can be replayed against the exact agent versions that produced it.

## What you can do

* **Run a research project.** One-shot brief in → multi-agent evidence pack + DecisionPacket out.
* **Run a survey.** A list of binary / multi / matrix / likert / star\_rating questions, each answered by a panel of agents with preserved disagreement.
* **Run a population test (Audience).** N=50-5000 personality-bound voter agents simulate how a real population would react.
* **Run a scenario pack (Simulation).** Monte Carlo over a factor space, producing fan charts + decision-point analysis.
* **Train your own agent.** Provide an interview transcript + corrections; the agent calibrates over time.

## Two ways to integrate

**For LLM-driven workflows** (Claude, ChatGPT, Cursor, your own agent) - use the MCP server:

```bash theme={null}
npx -y @rayify-ai/mcp
```

The MCP server exposes a slim 8-tool surface (4 pull + 4 push) plus 3 guided prompts. Designed for AI clients that need structured access to your Rayify workspace.

See [MCP Setup](/mcp-setup).

**For non-AI applications** (server backends, batch jobs, automation) - use one of the SDKs:

* Python: `pip install rayify-sdk`
* TypeScript: `npm install @rayify-ai/sdk`

See [SDK Setup](/sdk-setup).

## What's next

* [Quickstart](/quickstart) - your first project in 5 minutes
* [MCP Setup](/mcp-setup) - wire Rayify into Claude / Cursor / your AI client
* [SDK Setup](/sdk-setup) - Python + TypeScript SDK reference
* [API Reference](/api-reference) - the underlying REST API