FAQ

Answers about Ultralayer — what it is, how it works, and how to get started.

PRODUCT
What is Ultralayer?

Real-time market intelligence, built for your AI. We provide optimized financial context so your AI can analyze markets with better evidence, structure, and timing.

Who is Ultralayer for?

Retail investors using AI for analysis and decisions, and developers building AI-finance apps and agents.

How is this different from a generic LLM?

Ultralayer is not an LLM. We provide the real-time market intelligence and optimized financial context your AI needs to get better results — meant to integrate with ChatGPT, Claude, and other AI tools, not replace them.

How is this different from a traditional financial data API?

AI works with context, not raw data dumps. Ultralayer returns dense, scored, entity-grounded market context with citations — structured for agents and analysis workflows, not tables you have to reshape yourself.

PRODUCTS
What products does Ultralayer include?
  • The Wire — every market headline, classified and scored in seconds.
  • Developments & Events — structured market narratives with scored impact on tickers.
  • Filings Intelligence — official company guidance, revisions, and wording changes between earnings releases.
  • Semantic Search — find quoteable passages by meaning.
  • Stakeholder Identification — ranked public-company winners and losers with cited evidence.
  • Entity Retrieval — a full context package on any entity in one call.
  • Alerts — notify by email, Telegram, or webhook when a condition fires.
What is the Wire?

A novelty-clustered market headline feed. Each item includes entities, sector and asset tags, trust and importance scores, and links into storylines — so you see what’s moving without wading through syndication noise.

What are Developments and Events?

Developments are structured market facts with importance, surprise, and confidence scores, plus signed stakeholder impact and cited quotes. Events group related developments into longer-lived threads you can follow over time.

What is Filings Intelligence?

Official company outlooks from the press release and the earnings call — structured so you can compare companies, see every raise or cut, and score results against their own guide, not the street. Consecutive earnings releases are compared for what they added, rewrote, or quietly dropped.

What is Semantic Search for?

Retrieving quoteable document passages by meaning — from news, transcripts, and curated social — ranked for similarity, utility, and sentiment so your AI can ground answers in evidence.

What does Stakeholder Identification do?

Maps a market situation to ranked public-company winners and losers with verified tickers, signed impact scores, cited quotes, and a factor comparison matrix. It can take about one to two minutes per run because it runs a full retrieval-and-reasoning pipeline.

What does Entity Retrieval return?

A one-call dossier for a company, person, commodity, or country: identity, recent high-relevance headlines, a 14-day relevance-weighted sentiment view, and top co-mentions.

How do Alerts work?

You wrap an Ultralayer query, set a polling schedule (every 5 minutes or more), and choose delivery by email, Telegram, or webhook. When new matching results appear, we send them — with delivery-aware dedupe so a message means something new happened.

INTEGRATIONS
Where can I use Ultralayer?

In ChatGPT, Claude, Cursor, and other MCP clients; over the REST API; and in the web app. Same intelligence everywhere.

Do you support MCP?

Yes. Connect via our MCP endpoint, or install the Ultralayer plugin (MCP plus agent skills) where your host supports plugins. Auth is OAuth in the browser — no API key inside the AI client. MCP setup docs.

Do I need an API key for ChatGPT or Claude?

No. Plugin and MCP connections use OAuth sign-in. API keys are for the REST API and developer integrations.

Why doesn’t the ChatGPT / Claude / MCP widget show sources, quotes, or open links?

Usually because the agent called the tool with a reduced detail level (standard or essential). That omits heavier fields — source quotes, some provenance, scores — so the model uses less context. It is not a broken product or missing feature: the same endpoints return the full package when detail is full (the default). Ask the agent to re-run with full detail, or open the result in the web app. More on response detail.

PRICING & ACCOUNT
How does pricing work?

Pay-as-you-go credits — no subscription, no seats, no minimum. You pay for what you call. See pricing for current rates.

Do I get free credits to start?

Yes. New accounts receive a one-time grant of $0.50 in trial credits so you can evaluate the platform end to end before purchasing.

Do credits expire? What if a request fails?

Credits don’t expire. Failed requests are automatically refunded — you pay for results, not for attempts.

Are Alerts and MCP charged separately?

No. Alerts and MCP usage are billed at the underlying endpoint rate — no surcharge for scheduling or the MCP surface.

DATA & RIGOR
Where does the data come from?

Hundreds of channels around the clock — news and newswires, filings, earnings calls, press releases, and curated social, video, podcasts, and Telegram. Coverage is densest on US equities, especially large caps, with global market tape in the mix. More in our sources docs.

Do answers include sources and evidence?

Yes. Claims come with citations and quotes checked against retrieved source text. When something is an AI synthesis, it is labeled and sourced as one.

Did you measure Ultralayer against ChatGPT?

Yes, with a caveat. We did not use the ChatGPT web interface. We ran the same agent, openai/gpt-5.6-luna, through the API with OpenAI web search. That agent ran each task twice — once with web search, once with web search plus Ultralayer. Mean scores were higher with Ultralayer on completeness, timeliness, and insight. Method, results, and the full task set are in the benchmark report.

Can I backtest with the same queries I use in production?

Yes. The same endpoints serve realtime and point-in-time backtests, so the query you backtest is the query you deploy. Filters use knowledge-time so you don’t get lookahead leakage. Backtesting docs.

GETTING STARTED
How do I get started?

Create an account in the console. From there you can connect ChatGPT, Claude, or any MCP client; explore the app; or create an API key for developer integrations. Docs walk through each path.

Where can I get help?

Reach us on the contact page, or start with the docs.

START HERE
$0.50
in trial credits, free on signup
No card, no subscription, no minimum. Enough to test every endpoint end to end — around 160 live market scans, or 3 full stakeholder analyses.