For the complete documentation index, see llms.txt. This page is also available as Markdown.

Surfaces & analyst personas

One Clarity analyst answers every surface. What changes is the persona, the answer shape, and what it can see.

Every Clarity answer — a chat reply, a dashboard card, a daily client analysis — follows the same analyst approach. It draws on the same evidence and follows the same citation rules everywhere. Interactive answers run the research loop for the question in front of them; scheduled cards and analyses reuse daily global and stock research through compact cited handoffs before adding client-specific synthesis. What changes from one surface to the next is three things:

  • The persona it adopts — a single-stock analyst, a portfolio analyst, a top-down advisor, or the general analyst.

  • The answer shape — a streamed conversation, a fixed two-section card, or a saved daily write-up.

  • What it can see — whether the customer's actual holdings and the conversation history are in scope.

There is no separate store of market truth for a card or a chat. Pick a surface and you are choosing a persona, an answer shape, and how much saved daily context can be reused on top of the same evidence memory.

The personas below are roles the one analyst plays, not different products. Each maps onto a product: the single-stock persona powers asset insights, the portfolio persona powers portfolio insights, and the top-down advisor powers market insights.


One analyst, four personas

A persona changes which evidence the analyst leads with and how it structures the answer. It never changes the honesty rules: every persona still answers only from retrieved evidence, cites every fact, and refuses to invent the parts Clarity does not cover.

Persona
What it does
Leads with

General analyst

Open-ended research on any entity, theme, or "what's happening" question

Whatever the question is about

Single-stock analyst

A close read on one company or token

Fundamentals and price, then how the situation evolved, then the macro links and risks

Portfolio analyst

The customer's holdings as a whole

The single biggest development moving the portfolio, then each material position

Investment advisor

A top-down market read

Geopolitics → macro & policy → sector impact → positioning (beneficiaries, casualties, hedges)

In the admin panel's Analyze chat you can switch persona per question. Every other surface picks the persona that fits its job automatically.


Surface map

Surface
Who it's for
Persona
Shape
Sees the customer's holdings

Clarity chat — Analyze tab

Your team, in the admin panel

Any of the four (you pick)

Streamed conversation

No

Clarity chat — customer app

A signed-in portfolio client

Portfolio-aware general analyst

Streamed conversation

Yes — their real positions

Clarity chat — share link

Anyone with a public portfolio link

Portfolio-aware general analyst

Streamed conversation

Yes — the shared portfolio

Clarity chat — partner embed / API

A user inside your own app

Portfolio-aware general analyst

Streamed conversation

Yes — holdings you supply

Clarity card — portfolio

A portfolio dashboard

Portfolio read

Global context + portfolio impact + options to consider

Yes

Clarity card — single asset

A stock or token page

Single-stock analyst

One-shot commentary + recent news

No

Daily portfolio analysis

Each client, every day

Portfolio analyst

Saved write-up, advisor-reviewed

Yes

Daily holding analysis

Each position, every day

Single-stock analyst

Saved write-up, advisor-reviewed

Yes

Monitoring

Continuous, event-driven

Top-down + portfolio

Alerts and digests

Yes

A few distinctions worth holding onto:

  • Chat vs card vs analysis. Chat is conversational and streamed — the answer appears as it generates. A card has a fixed layout and is precomputed from the day's layered artifacts. A daily analysis is generated on a schedule and saved.

  • Remembering the conversation. The admin Analyze chat and the signed-in customer chat remember the thread — earlier turns, confirmed choices, and a running scratchpad carry forward. Share-link and partner-embed chats answer each call fresh, with continuity coming from the message history you send. Cards and daily analyses are single-shot.

  • The daily analyses carry your firm's endorsement. Clarity drafts a portfolio-level analysis and a per-holding analysis for every active client each day. They wait in the Review tab until an advisor approves or amends them, and each new day's read builds on the last one for continuity.

  • The partner surfaces own their holdings data. For the embed and API card and chat, you pass the portfolio snapshot with the request; Clarity treats it as the source of truth for positions rather than looking them up. See Integrate Clarity and Embed Clarity.


What the analyst can draw on

Whatever the surface, the analyst answers from a fixed set of evidence sources — and only from them.

  • Clarity's memory — the entities, facts, relationships, and story arcs built from ingested news, filings, transcripts, and uploads. This is Clarity's shared, global knowledge plus anything your organization has added on top. It is the backbone of every answer. See Clarity in depth.

  • Live market data — current price, recent change, performance over standard windows, and volatility for any company or token that resolves to a ticker. Prices are overlaid at the moment you ask and are never stored as facts, so commentary always reflects the live market without Clarity's knowledge going stale.

  • Company fundamentals — market cap, P/E, P/B, enterprise value, and sector for ticker-linked companies.

  • The customer's holdings — on a portfolio-aware surface, the analyst can read the actual positions (units, value, cost basis, profit and loss, sector) to answer "what do I own" and "how am I doing". Outside a portfolio session it simply doesn't have them, and says so. See Portfolio context.

  • Yesterday's analyses — each new position read can pull the prior successful compact position handoff, so it builds on the last view without replaying an unbounded history.

  • Daily artifact handoffs — scheduled global and stock research passes their thesis, causal links, risks, freshness, and citations into position and portfolio synthesis. Personalized prose and holdings stay outside Clarity's shared evidence memory.

  • The conversation and a scratchpad — in a remembered chat, the analyst sees the full thread and keeps a small working note of confirmed choices and open questions, so it doesn't re-ask what you already told it.

  • Source provenance — every fact traces back to the exact article, filing, or upload it came from, which is what each citation resolves to.


How it researches and answers

The analyst is a research-first agent. It does not answer from memory or the open internet — it gathers evidence with tools, then composes a cited answer. There is no fixed step budget; it takes as many research calls as the question needs.

Its toolkit, in plain terms:

  • Search Clarity's memory — find the entities a question is about, pull the individual facts about them, and discover the ongoing story arcs around a theme.

  • Run a deep retrieval — hand the whole question to Clarity and get back everything relevant: the entities, facts, causal links, stories, and live prices in one pass. This is the workhorse for "give me the read on X".

  • Follow structure and time — walk an entity's current connections (supply chains, roles, exposures), read its timeline to see what changed and what has been superseded, expand a story arc end to end, and trace any fact back to its original sources.

  • Pull market data — live price and performance, company fundamentals, and, for a theme or commodity with no direct ticker, a list of tradable proxies to choose from.

  • Read portfolio context — the customer's holdings and the prior days' analyses, on a portfolio-aware surface.

  • Close the turn — submit the final, cited answer.

The answer itself always follows one contract, on every surface:

  • A thesis up top. It opens with the "so what" in one or two plain sentences — what the evidence means for an investor and why it matters now — not a restatement of the question or a company description.

  • Cite or omit. Every material point — a number, a driver, a cause, a relationship, a state change — is traced to a specific retrieved fact, shown as an inline citation. If the analyst can't point to evidence, it drops the point or names the gap explicitly. An uncited assertion is treated as a fabrication, so the analyst would rather tell you what it doesn't cover than guess.

  • A positioning read. On every surface, the analyst names who benefits and who is exposed from the cited evidence, then adds a hedge or diversifier steer flagged as its own analytical inference — never a personalized buy or sell instruction.

  • Current state, not stale state. For "what's happening now" questions it reads the timeline before describing the present, so a fact that has since reversed isn't presented as current.

  • Plain language. Written so someone with no finance background follows every sentence, with any unavoidable term explained in the same breath.

  • Evidence-driven sections and follow-ups. The body is grouped under headers chosen to fit the question and the evidence, and a normal answer ends with a few suggested follow-up questions.

  • A staleness signal. Answers carry how recent their freshest source is, so you can flag when the underlying evidence is old.

For the deeper mechanics of how a question is routed and how retrieval is layered, see the Pipeline. For the structured story-thread view that ships alongside each answer — the arcs, what developed, and what changed — see Story threads behind the answer.


Choosing a surface

  • A conversational "ask Clarity" box in your product → the chat, or the no-code embed.

  • Portfolio commentary above the fold → the card: two short sections, follow-ups, citations, and a staleness badge.

  • A close read on one stock or token → the single-asset card, or ask the chat about that name.

  • A daily, advisor-vetted note per client and per holding → the daily analyses in the Review tab.

  • Always-on, "tell me when something changes"Market signals & monitoring.

  • Internal research and spot-checking → the Analyze chat in the admin panel.


Good to know

  • Asset domains gate every surface. Each organization controls whether Clarity may use equities, crypto tokens, or both, on the Settings → Clarity page. Turning a domain off removes its assets from every card, chat, and analysis. See Asset domains.

  • The analyst runs on the model you choose. Which AI model powers Clarity is set per organization on the AI providers page.

  • Cost is the same across personas. Every persona uses the same research loop and is billed the same way — only the answer shape differs. See Cost.


  • Integrate Clarity — call the card and chat from your backend

  • Embed Clarity — drop Clarity into a browser with no API key exposed

  • Portfolio context — the holdings model the portfolio-aware surfaces run against

  • Pipeline — how Clarity ingests evidence and answers questions

  • In the admin panel — explore, analyze, and review client analyses

  • Products — market, asset, and portfolio insights, proposals, and monitoring

Last updated