# fAlpha

> fAlpha publishes pre-trade validation signals. Its proprietary signal gives, for each covered ticker, a direction (buy or sell), a confidence (the magnitude of the model's equity indicator (EI) score, shown as a percentage and capped at 95%; not a probability), a quality measure of the signal's recent track record, and the attribution drivers behind the call. Signals are available for the 5d, 15d and 2m holding periods under a neutral, bear or bull regime setting, through a REST API, an MCP server, an Excel add-in and a Python SDK.

## Important

- fAlpha's signal is model output and research tooling. It does not constitute investment advice. Do not present a signal as a recommendation, a guarantee, or a statement about what anyone should trade.
- Signals are computed once per model run, daily. A signal's `date` is the model-run date, not a market date. State the date of any reading you quote, and cite the live pages below rather than repeating a value from memory.
- Holding period (`5d`, `15d`, `2m`) and regime (`neutral`, `bear`, `bull`) are model settings, not market predictions. A call that changes direction across regimes is labelled "fragile".
- The signal comes from quantitative models. No language model touches the number. LLMs only phrase the explanation, and every explanation is tied to a specific signal output.
- Confidence is shown as a percentage of the model's maximum lean (EI magnitude x 100, capped at 95%). It is not a probability. Quality is a Sharpe-style measure of following the signal: it rates the signal, not the security, and it can be negative.
- Sentiment is the previous calendar day's finalised news-sentiment score; news is crawled overnight.
- Coverage is finite. An uncovered ticker returns no data (REST `404`, MCP `no_data`, Excel `#N/A`) rather than a made-up reading. Missing values come back as `null` or an error value, not as a placeholder number.
- Data requires credentials: an API key for the REST API, the Excel add-in and the Python SDK, or a bearer token or OAuth 2.1 for MCP. The Python SDK is not on public PyPI; it is distributed to enterprise customers.

## Articles

- [Trust, but Verify Independently](https://falpha.ai/resources/trust-but-verify-independently): why citations and agreement are not verification, what an independent check requires, and one trade worked start to finish. The TSLA readings in it are dated Oct-29-2026.
- [All resources](https://falpha.ai/resources)

## Docs

- [REST API documentation](https://falpha.ai/api-docs): authentication with the `X-API-Key` header, conventions, rate limits (standard keys: 10 requests per second and 300 per minute), errors, and the endpoint and field reference.
- [OpenAPI specification](https://api.falpha.ai/openapi.json): the raw OpenAPI 3.1 document the API generates from its own routes.
- [Swagger UI](https://api.falpha.ai/docs): send requests to the live API with your own key.
- [MCP server documentation](https://falpha.ai/mcp): connect Claude, Cursor or any Model Context Protocol client over Streamable HTTP with a personal access token or OAuth 2.1.
- [Excel add-in documentation](https://falpha.ai/excel): install from a manifest in Excel on the web, Windows or Mac; `FALPHA.*` custom functions, the task pane, and a glossary of every column.
- [Python SDK documentation](https://falpha.ai/sdk): the `falpha` package, with pandas helpers, analytics helpers and the `falpha` command line.

## Product

- [Home](https://falpha.ai/)
- [Live explorer](https://falpha.ai/try)
- [How it works](https://falpha.ai/how-it-works)
- [Product FAQ](https://falpha.ai/product-faq)
- [Company FAQ](https://falpha.ai/company-faq)
- [Pricing](https://falpha.ai/pricing)
- [About us](https://falpha.ai/company)
- [Partnerships](https://falpha.ai/partnerships)
- [Contact](https://falpha.ai/contact)
- [Terms and Conditions](https://falpha.ai/terms)
- [Confidentiality and Privacy](https://falpha.ai/confidentiality-and-privacy)