Quickstart

There are two ways to start, depending on what you want to do:

They are the same platform, the same account, and the same credits — just two doors into it. Many people end up using both.

Use Ultralayer in your AI tools

Ultralayer ships as an MCP server, which means any MCP-capable tool — ChatGPT, Claude, Cursor, Codex, and others — can query the wire, run searches, analyze stakeholders, and manage alerts directly inside the conversation. For coding agents we also ship a plugin that bundles the MCP server together with agent skills: detailed instructions that teach your agent to use each capability well, without you having to prompt-engineer anything.

1. Create an account

2. Connect your tool

Wherever your tool asks for an MCP server URL, it is:

Code
https://api.ultralayer.ai/v0/mcp

When you connect, your tool opens a browser window to sign in with your Ultralayer account. No API key is needed for MCP — authentication is handled through OAuth.

3. Ask something

What's moving markets right now? Focus on high-importance items only.

Brief me on Nvidia: recent headlines, sentiment over the last two weeks, and who co-moves with it in the news.

What are the upcoming developments for Tesla — scheduled earnings, product launches, or other future events?

Set up an alert that emails me when Apple prints important news with strongly positive or negative tone.

Your tool will call the corresponding Ultralayer capability — The Wire, Entity Retrieval, Developments & Events, and Alerts for the examples above — and answer with sourced, structured results.

Build with the API

1. Create an account and an API key

2. Make your first request

Code
curl -X POST "https://api.ultralayer.ai/v0/search/list_wire" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "max_records": 10,
    "novelty": ["new", "update", "correction"],
    "min_importance_score": 0.5
  }'

Each returned item is a compact, structured object:

Code
{
  "wire_id": 67583,
  "headline": "Company X raises FY guidance to $4.2B on data center demand, shares up 8% premarket",
  "timestamp": "2026-07-23T13:02:11Z",
  "source_timestamp": "2026-07-23T13:01:45Z",
  "novelty": "new",
  "importance_score": 0.72,
  "trustworthiness_score": 0.85,
  "entities": [
    { "canonical_name": "Company X, Inc.", "entity_type": "COMPANY", "relevance": "primary", "sentiment": 0.6 }
  ],
  "about": { "sectors": ["information_technology"], "asset_classes": ["equities"] }
}

A few things worth noticing:

• novelty: ["new", "update", "correction"] filters out syndicated duplicates — you see each story once.
• Every item carries two timestamps: when the source published it and when it entered Ultralayer. This is what makes honest backtesting possible.
• The scores are there so you can threshold the firehose instead of reading it.

3. Go deeper

• Each capability page documents its endpoints with use cases and copy-paste request bodies: The Wire, Semantic Search, Developments & Events, Stakeholder Identification, Entity Retrieval, Alerts.
• The API Reference has the full schema for every endpoint, with schema download available.
• Credits & billing explains how usage is metered.

Where next

• Ways to use Ultralayer — a map of every surface: app, API, MCP, plugin, console.
• Concepts — how point-in-time integrity, sourcing, and scoring work under the hood.