> For the complete documentation index, see [llms.txt](https://docs.uprightplatform.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.uprightplatform.com/upright-agent/data-and-tools.md).

# Data and tools available

The chat does not produce numbers from a generic large language model. Every quantitative answer it gives comes from a structured tool call against Upright's own data — the same data that powers the rest of the platform. This page lists what the chat can reach and how it routes between sources.

## Analytical lenses

The platform exposes six analytical lenses for a company, each with its own dedicated chat tool. The chat never reconstructs one lens's answer from another lens's data.

* [**Net impact**](/metrics/net-impact.md) — Upright's headline impact score, broken down by category and dimension. Triggered by phrases like "net impact", "impact ratio", or impact category names (GHG emissions, jobs, knowledge, …).
* **CSRD Double Materiality Assessment** — material topics under ESRS, plus drivers and supporting evidence. Triggered by phrases like "DMA", "double materiality", or ESRS matter names ("climate change mitigation", "own workforce", "biodiversity", …).
* **Financial Effects** — monetary translation of sustainability risks and opportunities. Triggered by "financial effects", "EUR risk size", or "profit impact". Available on specific releases only. Beyond reading the headline figures, the chat can also size the financial effect of a single risk or opportunity you describe: it first echoes back the inputs it will use for you to confirm, then returns the monetised effect in the company's reporting currency.
* [**UN SDG alignment**](/metrics/un-sdg-alignment.md) — alignment of a company's revenue with the UN Sustainable Development Goals, down to the individual SDG-target level, with per-product reasoning for why a product supports or undermines a given target.
* [**EU Taxonomy**](/metrics/eu-taxonomy.md) — taxonomy eligibility and alignment, including green CapEx where available.
* [**SFDR Principal Adverse Impacts**](/metrics/sfdr-pai-indicators.md) — PAI indicators for sustainability-related disclosures.

Each lens has a dedicated platform deep link — when the chat answers from a lens, the first mention of the company (or topic) links straight to that lens's view in the platform.

## Single company, multi-company, group, and portfolio

The chat keeps companies, groups, and portfolios as separate first-class concepts and picks the right tool for the right scope.

* **Single company** — net impact, DMA, financial effects, SDG, EU Taxonomy, SFDR PAI for one named company.
* **Ad-hoc list** — when you paste a few names ("compare Apple, Microsoft, and Alphabet"), the chat returns per-company rows side-by-side. Ad-hoc lists do not produce a list-level aggregate; for a portfolio-level aggregate use a real group or portfolio that lives in the platform.
* **Group / portfolio** — when the platform already has a stored group (for example, an ETF, an index, a customer fund), the chat uses the group-level tools that return the platform's weighted roll-up, the holdings list, and the top contributors driving each impact category.
* **Screening** — for ranked or filtered lists ("the top 10 European utilities by net impact"), the chat uses Upright's screen tool rather than guessing constituents.

## Coverage — which entities the chat can answer about

The chat can answer about any company, fund, or product in Upright's coverage — on the order of tens of thousands of companies and funds and over a hundred thousand products. When you ask about something that is not covered, the chat says so rather than guessing, and (where relevant) can identify a company on demand from the public web. For the authoritative scope and how coverage is built, see the [Coverage](/coverage/off-the-shelf-coverage.md) section.

## External context: uploaded files and the web

Beyond Upright's own data, the chat can also reason over two external inputs you control from the composer (see [Using the chat](/upright-agent/using-the-chat.md#attaching-files-and-searching-the-web)):

* **Uploaded files** — a document you attach (for example, your own materiality assessment) that the chat reads alongside Upright's results.
* **Web search** — public-web context the chat pulls in only when you turn the **Web search** toggle on. Web-sourced claims are cited like any other source.

## Supporting evidence and drill-downs

Beyond headline numbers, the chat can answer "why" questions by pulling in:

* **Product-level reasoning** — why a specific product is assessed positively or negatively on a given impact.
* **Supporting articles** — the scientific evidence Upright uses to ground its product impacts.
* **Contributing products** — the products driving a company's score on a given impact category.
* **DMA reasoning, overrides, and key takeaways** — the platform's own explanations for why a topic is or is not material.
* **Single-matter summary** — drivers, IROs (impacts, risks, or opportunities), severity, and time horizon for one ESRS matter.

## Product value chain

The chat can render Upright's product taxonomy as an upstream / downstream value-chain diagram (for example, "what feeds into steel" or "what does steel feed into"). The diagram appears as a card below the chat answer; the prose narrates the tiers and quotes the share-of-value-add percentages.

## Site identification and physical climate risk

For questions like "where are X's offices / data centres / plants" the chat uses a public-web pipeline to identify site locations on demand. Combined with the climate-risk tools, it can also answer "is X exposed to wildfires / heat / heavy precipitation at its sites", for individual locations or aggregated across all of a company's sites.

For broader spatial questions ("show climate risk across this region") the chat can render a grid-level climate-risk heatmap as a card below the answer, selectable by hazard, scenario, and time horizon.

## Knowledge base and methodology

The chat has direct access to this knowledge base. Methodology questions ("how does Upright weight impact categories", "what does the SFDR PAI alignment column mean") are answered from the knowledge base content via the chat's search tool, with citations back to the relevant page.

## Comparisons across releases

The customer-facing chat is scoped to Upright's **latest** model release. Comparisons against older releases ("how did the 4.x DMA differ from now") are not supported on the chat surface — they require direct platform access.

## What the chat never does

* **Never fabricates numbers** for a lens that has no data for a given company. If SDG / EU Taxonomy / SFDR PAI data is not available for a company, the chat says so rather than estimating from net impact or DMA figures.
* **Never reconstructs one lens from another** — for example, it does not produce SDG alignment by re-deriving it from net impact, or estimate DMA materiality from SDG alignment.
* **Never invents methodology coefficients** — for example, the financial-effects formula is multiplication of three named factors with no hidden weighting coefficients, and the chat will not invent any.