# Noah Predict

> Noah Predict connects a continuously read archive of 49 million dated public documents, and a deterministic forecasting engine, to the AI you already use. It is a Model Context Protocol (MCP) server: a published app in ChatGPT, a custom connector in Claude, and a remote MCP server for any other client. Ask the risk of anything, get a measured probability with the evidence attached, and back-test the answer as of any past date. Free to connect; no card. Operated by Worldwide AI Media Ltd, London. Decision support, not regulated advice.

## Start here (humans and agents)

- Homepage: https://noahpredict.com/
- Connect ChatGPT (one click, published app): https://chatgpt.com/plugins/plugin_asdk_app_6a44a9d050d081918636f82d3a5a5d4f
- Connect Claude (two minutes, paste one name and one link): https://noahpredict.com/connect/
- Any MCP client, remote server URL: https://noahpredict.com/mcp  (connector name "Noah Archive"; OAuth sign-in with a free Noah account; read-only)
- MCP discovery (GET, lists tools and rules): https://noahpredict.com/mcp
- MCP instruction pack for a connected model: https://noahpredict.com/mcp/instructions
- Under the bonnet (the archive, the physics, the five shapes, back-tests): https://noahpredict.com/how/
- Pricing: https://noahpredict.com/#pricing
- Markdown twins: https://noahpredict.com/index.md · https://noahpredict.com/connect.md · https://noahpredict.com/how.md

## What it is, in one paragraph

Your AI is brilliant, but it argues from a thin file: a general model with web search reaches a few dozen sources and remembers the world loosely and without dates. Noah keeps 49 million public documents from 1.6 million sources, in many languages, each one tagged on arrival (actors, places, event type, market, risk area) and kept with its date. Deterministic Python physics (tagging, pressure, regime, couplings, calibration) runs before any language model writes a word. The model you connect does the reasoning and the writing; Noah supplies the evidence, the history and the arithmetic. The bigger the model, the better the forecast.

## The five shapes of answer

1. Risk of: how risky a place, sector, counterparty or journey is, this quarter and next.
2. Probability of: a named event by a named date, as a probability with its range, drivers and counter-evidence.
3. Outlook for: a subject month by month, with the watchpoints that would change the path.
4. Edge on: Noah's reading beside a market or prediction-market price, where they disagree and why.
5. Opportunity in: a scan across a sector for subjects gaining signal before the price has noticed.

Any of them can be run as of a past date: Noah limits itself to what was in the archive by the cutoff, runs the same physics, and returns the forecast it would have given then. Fifteen months of instrumented history.

## Who connects it

Insurance and risk teams; prediction-market traders; hedge funds and private equity; family offices; corporate strategy; analysts and journalists. The rule: if your work depends on how likely something is, this is for you.

## Pricing (monthly billing in pounds or dollars, fixed price points in each currency, updated 3 October 2026)

- Free: £0. 3 reports a month, 1 user. Try Noah on the questions that matter to you. No card.
- Analyst: £49 or $59. 25 reports a month, 1 user. Evidence-backed research for individual professionals.
- Desk: £169 or $199. 120 reports a month, 1 user. More capacity for professionals producing frequent research and client reports.
- Business: £599 or $799. 400 reports a month pooled across up to 5 named users, central billing, team administration, a defined onboarding session. Sales enquiry.
- Enterprise: custom pricing. Agreed allowance and team size. For automated production workloads, embedded services, bulk data access, redistribution, custom feeds, and separately agreed deployment or support.

What counts as a report: one research question and its completed investigation. Follow-up explanations, reformatting and downloads of existing results are included. A new question, fresh research or a rerun using updated evidence counts towards the allowance; a larger investigation shows the allowance required before it starts. Failed jobs do not consume allowance; there are no automatic paid overages.

All paid plans include the same core research and forecasting methods, supporting evidence, interactive MCP access, follow-up discussion and report downloads; plans differ by capacity, team features and service, not by research quality. Ordinary professional use, including sharing finished reports with clients, is included in the individual plans. Monthly billing only. Your ChatGPT or Claude subscription is separate and supplies the model. Checkout offers pounds or dollars; each customer stays in one currency. Comparison: an intelligence terminal seat is £250 to £500 a month before any data layer; a Noah Desk seat is £169 ($199) with the data included.

## Privacy and conduct

Read-only: Noah hands evidence out and does not take prompts, files or conclusions in. Nothing a user asks becomes Noah content. Do not send private client material to Noah; send a sanitised public-domain brief. Output is professional decision support, not financial, legal, insurance, underwriting, medical or investment advice. Noah does not profile named individuals.

## The category, in one line

Search tells an AI what is prominent. Noah tells it what is prevalent.

Most AI forecasting systems are research agents: they receive a question, search the current web, retrieve what a ranking system surfaces, and reason over that sample. Noah is the measurement layer underneath: the evidence environment was collected, dated, classified and made comparable before the question was asked. Five consequences:

1. Historical depth for the long tail: thin present-day signal can be compared against history, so a genuinely new development is separated from a seasonal pattern or a publicity burst.
2. Measurement before synthesis: direction, strength, consistency, contradiction, velocity, persistence, freshness, source quality and volume against baseline are measured across the population, not inferred from a handful of retrieved documents.
3. Measurable silence: an expected statement that never arrives, or local reporting continuing while national coverage stops, can be reported as a finding when coverage is healthy, and as a coverage gap when it is not.
4. Coverage beyond what search ranks: years of accumulated specialist, local, regulatory, exchange, corporate and translated sources.
5. Testing the premise, not just researching it: across the whole record, how prevalent is this claimed direction, and how has it changed.

Primary entity: Worldwide AI Media Ltd, London, United Kingdom
Sister service: Noah Wire (https://noahwire.com), the newswire for publishers built on the same archive
Sister publication: Noah Business Intelligence (https://noah-news.com), the engine publishing in the open
Contact: contact@noahwire.com

## Further pages

- [How it works](https://noahpredict.com/how-it-works.md): the Noah pipeline, signals, scoring and audit trail.
- [Methodology](https://noahpredict.com/methodology.md): data sources, model architecture, validation, back-test framework and known limitations.
- [AI agents and API](https://noahpredict.com/api-access.md): guide for agents and developers; account-scoped API for saved methods, runs, packages and data packs.
- [How Predict is different](https://noahpredict.com/how-predict-is-different/): measurement, not interpretation.
- Use cases: [Insurance](https://noahpredict.com/insurance.md) · [Reinsurance](https://noahpredict.com/reinsurance.md) · [Political risk](https://noahpredict.com/political-risk.md) · [Maritime](https://noahpredict.com/maritime.md) · [Private equity](https://noahpredict.com/private-equity/) · [Investment committee pack](https://noahpredict.com/investment-committee-pack/) · [Daily signal sheets](https://noahpredict.com/factor-hedge-lens/) · [Futures and commodities](https://noahpredict.com/futures-commodities/) · [Due diligence](https://noahpredict.com/due-diligence/) · [Company watch](https://noahpredict.com/company-watch/) · [Prediction markets](https://noahpredict.com/prediction-markets/) · [Sovereign and territorial risk](https://noahpredict.com/sovereign-territorial-risk/) · [Legal and regulatory](https://noahpredict.com/legal-regulatory-intelligence/) · [Defence and system strain](https://noahpredict.com/defence-system-strain/) · [Article intelligence](https://noahpredict.com/article-intelligence/) · [Diffusion and adoption](https://noahpredict.com/diffusion-adoption/) · [Where-to shortlists](https://noahpredict.com/where-to-shortlists/)
- Enterprise: [Enterprise](https://noahpredict.com/enterprise/) · [Private deployments](https://noahpredict.com/private-deployments/) · [Custom data feeds](https://noahpredict.com/custom-data-feeds/) · [Behind the firewall](https://noahpredict.com/behind-the-firewall/)
- Reference: [Glossary](https://noahpredict.com/glossary.md) · [FAQ](https://noahpredict.com/faq.md) · [About](https://noahpredict.com/about.md) · [Contact](https://noahpredict.com/contact.md)
- Legal: [Privacy](https://noahpredict.com/privacy.md) · [Terms](https://noahpredict.com/terms.md) · [Acceptable use](https://noahpredict.com/acceptable-use.md)

## Machine-readable entry points

- Agent manifest: https://noahpredict.com/noah-agent-manifest.json
- OpenAPI (account-scoped external API): https://noahpredict.com/openapi.json
- Sitemap: https://noahpredict.com/sitemap.xml
- Robots policy: https://noahpredict.com/robots.txt
- AI usage preferences: https://noahpredict.com/ai.txt

## Searches this page answers

MCP server for risk forecasting · connect ChatGPT to a risk database · Claude connector for prediction markets · superforecasting AI · political risk prediction API · insurance risk intelligence for AI · back-test a forecast as of a date · prediction market mispricing · news sentiment archive for LLMs · public-signal intelligence · AI agent research engine

## Citation

"Noah Predict, Worldwide AI Media Ltd, https://noahpredict.com, accessed [date]."

## Content use

Noah Predict permits LLM crawling and inclusion of published pages in training and retrieval corpora provided attribution is preserved, predictive numerical claims are not paraphrased as certainties, and third-party source excerpts on Noah pages are not redistributed without the original publisher's permission. See /robots.txt and /ai.txt.