Every question, answered.

No. Profiles are sent to your assistant as context at the moment you ask a question, in the same way any MCP tool returns data. Anthropic and OpenAI both offer commercial terms that exclude API and connector traffic from training, and those are the terms we operate under.

We do not train anything of our own on your team’s data either — not a shared model, not an aggregate benchmark, not an "anonymised" corpus.

It sits either side of the decision. Before: sharper, more structured interviews and a debrief with precise language. After: an offer shaped around what the person values, and a manager who knows how to work with them from day one.

Your scorecards, work samples and interview loop stay exactly as they are. Throughline adds the part most processes skip — what happens once they say yes.

36 statements, about 7 minutes. Your progress saves on your device, so you can safely come back to your link. You'll see your full result immediately — nothing is shared with your team until you confirm it.

Two, for pair guidance — which is where most of the value sits. The team map needs about four before its observations mean anything, because below that it's describing a coincidence rather than a shape.

You don't need full coverage, and pushing for it tends to produce compliance rather than consent.

Four measures, and the first is the honest one: unprompted queries per manager per week, after the novelty has gone. Then a two-question pulse before and after, one-to-one specificity sampled by the managers themselves, and time-to-productive for new joiners.

The metric to distrust is opt-in rate. It's easy to move by pushing and it tells you nothing about whether anyone found it useful.

Not usually, and pretending otherwise would cost you more than it costs us.

Predictive Index is built to support decisions about people — selection, with the job analysis and adverse-impact work that requires. We are built to support conversations with them. If you need defensible selection testing, buy that; we have deliberately not built it.

Where we do replace something is the annual workshop whose output nobody opened again.

No. Dominance describes a preference for directness and speed, not hostility. A Dominance colleague states positions plainly and is comfortable with disagreement, which people who prefer more social cushioning often experience as aggression. The distinction matters practically: aggression is about intent to dominate a person, while the Dominance style is about impatience with anything between a problem and its resolution. Most of the friction attributed to aggression is a mismatch in how much cushioning each person expects.

No. Steadiness describes a preference for continuity and relationships over confrontation, not an absence of conviction. A Steadiness colleague frequently holds firm positions and simply weighs the cost of asserting them differently — the calculation is about what the disagreement will do to the working relationship, not about whether they have a view. Reading Steadiness as passivity is the most common way managers end up surprised by a resignation from someone they thought was content.

Not quite, though they overlap. Extraversion is a Big Five trait measuring where someone draws energy from and how much external stimulation they seek. Influence is a working style derived from high extraversion combined with high agreeableness — the energy is outward and it is oriented toward people rather than toward the task. A high-extraversion, low-agreeableness colleague reads as Dominance rather than Influence, despite being equally sociable.

No. Profiles are sent to your assistant as context at the moment you ask a question, exactly as any MCP tool returns data. Anthropic and OpenAI both offer commercial terms that exclude API and connector traffic from training, and those are the terms we operate under.

We do not train anything of our own on your team's data either — not a shared model, not an aggregate benchmark, not an anonymised corpus.

No, and the shared name causes real confusion. The Big Five trait of conscientiousness measures how organised and follow-through-oriented a person is, and it is the strongest single trait predictor of job performance. The Conscientiousness working style is derived from a combination of lower extraversion and lower agreeableness, and describes a preference for precision and reserve. Someone can score high on the trait and present as a different working style entirely.

It would, if we shipped labels. Throughline records continuous scores, so two people who share a headline style still get materially different guidance — and the profile records where behaviour flips with context: quiet in a senior room, expansive with familiar peers.

The practical guard is the swap test. If a piece of advice would read just as sensibly with a different colleague's name in it, it has failed and it doesn't ship.

The Big Five personality traits are openness to experience, conscientiousness, extraversion, agreeableness and emotional stability — often remembered as OCEAN, with the fifth inverted as neuroticism. Each is a continuous scale rather than a category, so a person has a position on all five rather than belonging to a type.

The four working styles are Dominance, Influence, Steadiness and Conscientiousness. Dominance is direct and decisive and wants the conclusion first. Influence is animated and exploratory and thinks out loud. Steadiness is relational and consistent and needs the team with it. Conscientiousness is considered and precise and wants time before committing. Each describes how a person works, not how well.

Four things, all of them readable by your assistant rather than filed somewhere.

A profile for each person, in plain language — how they take feedback, make decisions and behave under pressure. A reading for every pair, naming where the friction comes from and what each side should do. A read on the whole team, including what it systematically misses. And ready structures for one-to-ones, critical feedback and delegation, fitted to the specific person.

Three situations come up more than the rest.

Writing feedback or preparing a one-to-one, where the guidance is about the person receiving it rather than about feedback in general. Naming a working relationship that has been quietly grinding for months, where neither person can explain it from their own results. And making a personality workshop actually stick, by moving the insight to where the work happens.

It measures how each person on your team works — using the Big Five, a working-styles assessment and the Enneagram — and makes those measured profiles available inside Claude, ChatGPT or any assistant that speaks MCP.

So when you ask your assistant how to open a difficult conversation with a specific colleague, it answers about that colleague, citing what it read, instead of describing people in general.

Twenty minutes, live, on a demo team. Bring the trickiest working relationship you manage and we'll ask it in front of you. If the answer isn't visibly better than what you'd get from a generic assistant, we'd rather you didn't buy.

Two profiles tell you about two people. A pair reading is about the relationship between them — where it works, where it slips, and what each side should actually do differently.

In the product the pair is measured rather than assembled from two type descriptions. That is the difference that matters: the guidance names those two people, and it stops making sense if you swap the names over. The pairings written up here are the general cases, useful for learning the vocabulary.

No. There is no profile to return and nothing to infer from. The assistant reports that the person hasn't joined, and declines to speculate — which is the behaviour we test for on every release, because an assistant's instinct is to be helpful and improvise.

No. A colleague of any seniority sees the working-style summary you chose to share and the pair guidance for the two of you. Trait percentiles, individual answers and your guidance about other colleagues are visible only to you. There is no seniority override, because the moment one exists the consent stops meaning anything.

Yes, slowly. Traits are stable over months and drift over years, and the visible style shifts faster than the underlying trait does because context pulls behaviour around — a new role, a new manager, or a period of pressure can all move how someone shows up without much changing underneath. A profile taken during a reorganisation reads differently a year later, which is why re-testing should be free.

They already know — they opted in themselves and can see what they shared. Sharing is mutual by construction: you can see how to work with someone only if they can see the same about you, which includes your own reports seeing how to work with you.

Managers who treat it as a secret advantage tend to find it stops working, because the value is in the conversation it makes easier, not in the information asymmetry.

Yes. Sharing is mutual by construction — a colleague can see how to work with you only if you can see the same about them.

That is deliberate rather than a technical limit. A one-way window into how to handle somebody is a different product, and not one we want to have built.

No, and that is deliberate. Working style describes how someone communicates, not how well they will do a job, so scoring candidates on it would be poor measurement and, in most places, a legal risk.

Selection testing is a different discipline with a much higher bar — job analysis, criterion validity, adverse-impact monitoring. Predictive Index and Hogan have built that; we have built something else. What we do instead is make the hire work once you have made it.

Give them a criterion, not a deadline. “We decide Thursday regardless” produces anxiety and a padded answer, because the pressure has not reduced the uncertainty that is causing the delay. “What would you need to know to be comfortable at eighty per cent?” produces a short list, which is usually obtainable, and the decision then arrives on time and is honest.

Ask a specific question rather than an open one, ask it in private or in writing rather than in front of the group, and leave the pause. “Any concerns?” reliably returns nothing. “What is the part of this plan you would change if it were yours?” returns something real. The other thing that works is asking before the decision closes rather than after, because once a direction is set a Steadiness colleague will generally get behind it rather than reopen it.

Give them outcomes and constraints rather than methods, put the decision at the top of every conversation, and be direct when something needs to change. They will not resent a plain correction, and they will disengage from a hedged one. The harder part of managing Dominance is the relational damage they do without noticing — the useful intervention is naming the specific cost quickly, rather than asking them in general terms to slow down.

Give them a sentence to use rather than asking them to be more realistic. “I want to, let me check what it displaces and come back tomorrow” works because it does not require them to say no in the moment, which is the part they find difficult. Asking an Influence colleague to become someone who declines easily is asking them to stop being the person whose willingness you rely on the rest of the time.

About twenty minutes per person for the three tests, and four minutes once to connect your assistant. Pair guidance is available as soon as two people have finished; the team map starts meaning something at around four.

For measurement, the Big Five has the strongest evidence: better reliability, replication across cultures, and the clearest links to workplace outcomes. MBTI has poor test–retest reliability at the type level — many people are typed differently weeks later — which is why Throughline doesn't use it.

DISC-style working styles are a different kind of tool. They describe behaviour at work in a language a whole team can share, and they map onto Big Five traits in predictable ways. Throughline uses the Big Five to measure and working styles to talk about it — and runs both, so each checks the other.

Because nobody opens a drawer in the middle of writing a message. The away-day insight was usually real; the failure was one of location, not of content.

Putting the same measured profiles inside the assistant your team already works in means the question gets asked at the moment it matters — while a manager is deciding what to write, not six weeks earlier in a room with sticky notes.

Because the last one probably failed for two reasons, and both are structural rather than a matter of picking a better instrument.

It was about one person, and the expensive problems happen between people — no individual report can see friction that lives in neither profile on its own. And it was static: you cannot ask a PDF how any of it applies to the message you are about to send.

Throughline keeps the measurement and changes where it lives. The same insight is available mid-message rather than mid-workshop.

That it's fine, that nothing about their working life changes, and that there's no partial record. Then mean it — the programme's credibility rests on the first person who declines and experiences no consequence.

Consent given under pressure isn't consent, and the measurement gets worse as well as the ethics: people answer as the employee they think is wanted.

Their profile deactivates and every pair guide involving them goes with it, immediately. The team read recomposes around who's left rather than keeping a ghost in the shape.

Then you're probably right. Mistyping is the Enneagram's best-known real-world problem, which is why we show your full nine-score profile, tell you plainly when it's a close call, and highlight the runner-up type. Read both, see which one describes the why of your behaviour (not just the behaviour), and retake or reflect. You're the final judge — the questionnaire is a mirror, not an oracle.

Because you will find out either way, and finding out in month four is worse for you and worse for us.

Each comparison is written the way we would want one written about us: what the other tool genuinely does better, then the table, then who should buy which. If the honest answer is that you should buy the other one, we would rather say it on the call.

Yes, slowly. Traits are stable over months and drift over years, with a well-documented population trend: conscientiousness and agreeableness tend to rise with age, while openness and extraversion tend to ease. Life events, a role change and a sustained period of pressure can all shift a score. A profile from three years ago is a historical document, not a current one.

That is one of the things it is best at, with one honest limit.

A pair reading names the friction in language neither person hears as an accusation, and it tends to produce a working agreement rather than a verdict about someone's character. Two capable colleagues grinding for six months usually have a pace or directness mismatch that neither individual report could show them.

The limit: it describes how two people work, not who was right. If the problem is conduct rather than style, this is the wrong tool and it will say so.

Each person sees the guides for pairs they're in, and nothing else. You can't read how two of your colleagues work together — that's their relationship, not yours.

Yes. The free tier is a real team of five, indefinitely, with no card and no expiry. Most organisations get through the whole first stage before anyone needs to raise a purchase order.

Two people is enough to get pair guidance, which is where most of the value sits. The team map needs about four before its observations mean much. You do not need full coverage, and pushing for it tends to produce compliance rather than consent.

In part, and we'd rather say so than pretend. The nine types correlate with measured Big Five traits in the directions the theory predicts, and that finding has been replicated. The wings and the stress and growth lines have much less evidence behind them, so Throughline presents those as prompts for reflection rather than predictions.

That's also why your Enneagram result sits alongside your Big Five and working-style results rather than standing alone: the three are read together, and your Big Five acts as a cross-check on the type.

Often you do not have to choose. If your existing results are still trusted, the question is whether anyone is using them week to week — and that is an honest thing to measure before spending anything.

Start with the free plan: one team of up to five people, no card and no expiry. If managers are not reaching for it unprompted by week three, you have your answer at no cost.

The summary you chose to share, and the pair guidance for the two of you. Not your raw trait scores, not your individual answers, and not anything about your relationships with other colleagues. Sharing is mutual: if you can see how to work with them, they can see the same about you.

Your admin can see who has an active profile, because they are billed per active person. They cannot see a reason, a decline event or a date. To the assistant you are simply a person without a profile, and it declines to speculate about you rather than filling the gap. There is no shadow profile and no partial record.

You get pair guidance for the people who did and nothing at all for the people who didn't. There's no partial record and no shadow profile built from their calendar or their writing.

Two people is enough to be useful. Pushing for full coverage tends to produce compliance rather than consent, and a coerced assessment is poor data as well as a bad idea.

They still read differently, because every score sits on a continuum rather than in a box. Two people who share a label can sit a long way apart on directness or pace, and the guidance follows the scores rather than the name.

Shared-style pairs have their own failure mode worth knowing: they rarely produce friction, and they reliably produce blind spots. Two fast deciders will not slow each other down.

Conscientiousness tends to fit work where accuracy compounds and errors are expensive — engineering, research, finance, legal, quality, analysis. But this describes where the style is comfortable, not who will perform. Working style is not aptitude, and there is no defensible evidence it predicts success in a given role. Throughline refuses questions that use working style to support a hiring or promotion decision.

Dominance tends to fit roles with clear ownership, visible outcomes and genuine decision rights — running a function, leading a turnaround, early-stage operating roles, sales leadership. But this is a description of where the style feels comfortable, not a prediction of who will succeed. Working style is not aptitude, and there is no defensible evidence that it predicts performance in a given role. Throughline refuses questions that use style to make a hiring or promotion decision.

Steadiness tends to fit work where continuity, trust and sustained relationships matter — client success, operations, people functions, long-cycle delivery, roles holding a team or a service together. But this describes where the style is comfortable, not who will perform well. Working style is not aptitude, and there is no defensible evidence it predicts success in a given role. Throughline refuses questions that use working style for hiring or promotion decisions.

Influence tends to fit work with variety, visible people contact and an early-stage problem to shape — business development, product discovery, partnerships, client-facing consulting, founding roles. But this describes where the style is comfortable, not who will perform. Working style is not aptitude, and there is no defensible evidence it predicts success in a specific role. Throughline refuses questions that use style for hiring or promotion decisions.

Most teams find it lands best after the offer, as part of onboarding — which is where the value is anyway. The new joiner gets their own results first, and their manager gets a running start.

Taking it earlier is fine too, as long as it genuinely doesn't affect the outcome and the candidate is told that plainly.

None of them, and the research is fairly clear on this. Leadership populations over-represent Dominance, which is a fact about who gets promoted rather than who leads well. Each style has a characteristic leadership failure: Dominance decides without buy-in, Influence starts more than it finishes, Steadiness avoids the necessary conflict, Conscientiousness delays past the point of usefulness. Knowing your own is worth more than having a particular one.

Managers who have to write the feedback, run the one-to-one and delegate the work. Teams who want to name a friction rather than live with it. People leaders rolling this out across an organisation with consent intact. And hiring teams, who use it after the decision rather than to make it.

It is built for people who have to defend the choice of instrument to somebody — which is why the reasoning is published rather than asserted.

You can, and it will produce something fluent. It will also invent the parts it doesn't know, treat a two-point gap as a settled fact, and give you advice that would read identically for a colleague with the opposite profile.

What Throughline adds is measured scores for real consenting people, the relationship layer between any two of them, and rules about what may not be inferred.

Not on a schedule. Traits are stable enough over a few years that annual re-testing would mostly measure noise, and charging you for that would be hard to justify.

Retake when something real has changed — a new role, a significant life event, or a result the person genuinely does not recognise. A retake within the first month of taking a test is free, so a bad day does not cost you a credit.

If it speaks MCP, yes. We test Claude, ChatGPT, Cursor and Claude Code on each release; other clients work but we won't claim we've verified them. Enterprise plans can pin the server to a region or run it inside your own boundary.

They're three layers of one profile: the Big Five measures your underlying traits, DISC names the style colleagues see, and the Enneagram asks what's driving both. Throughline reads them together — where they agree, guidance gets sharper; where they disagree, it flags the tension as something worth exploring, never as an error in you.

Yes, for what it is designed to describe. Well-built four-style assessments are reliable — people get consistent results on a retest — and the four styles map onto well-established personality traits, particularly extraversion and agreeableness. Where researchers debate is whether four categories are the best way to carve up behaviour that really sits on continuous scales.

Throughline answers that by keeping the scores behind the letters and running the Big Five alongside, so the working-style read is cross-checked rather than standing alone.

Neither, for four of the five. Each end carries advantages and costs: high agreeableness builds trust and struggles to deliver hard news; low agreeableness surfaces problems early and is read as difficult. Conscientiousness is the partial exception, with a consistent positive relationship to job performance — though very high conscientiousness has its own costs in perfectionism and inflexibility.

Yes — one team of up to five people, indefinitely, with no card and no countdown. It's a real team, not a trial, so you can find out whether managers reach for it unprompted in week three, which is the only signal that matters.

“That's just how I am” is the failure mode of every personality tool, and the guardrail is that Throughline never explains a person's behaviour to someone else as a reason it can't change. Pair guidance is written as two sets of moves, one per person — both sides get something to do.

It usually means something is wrong and they have decided raising it would make things harder for other people. Going quiet is the Steadiness stress response, and because it looks like coping it tends to be noticed late. The productive move is a specific question in a private setting — asking what the hardest part of their week has been will get an answer where asking how they are will not.

This is where it is most useful. The profile already exists with consent, the person is staying either way, and the receiving manager starts knowing how to brief them instead of spending a quarter working it out.

We won't build you a ranking view, a promotion-readiness score, or an integration that infers profiles from calendars or messages. We also won't run a mandatory rollout — if the plan is to require it, we're the wrong supplier and the data would be worse anyway.

Because they are still forming a view, and offering an unformed one feels to them like a failure of the standard they hold themselves to. This is not disengagement, and treating it as such is the most reliable way to lose the contribution of the person most likely to have spotted the flaw. Send material in advance and ask for a written response afterwards, and the input appears.

Usually because in their model, someone with an opinion states it. They are not withholding an invitation, they are assuming one is unnecessary. This is the single most common Dominance misreading, and it is cheap to fix: say your view unprompted and early, before the decision closes. A Dominance colleague will engage seriously with a position offered directly, and will not think less of you for volunteering it.

Usually they have not changed their mind, because they had not yet made it up. Influence colleagues form views in the act of articulating them, so the first few versions are exploratory, not settled. What looks like inconsistency is a thinking process happening in public. The fix is cheap and works immediately: ask “is that a decision or are you still turning it over?” They will answer accurately.

Two reasons, and the second is the one people find more persuasive.

Selection testing is a different discipline with a much higher bar — job analysis, criterion validity, adverse-impact monitoring, and defensibility when a rejected candidate challenges the decision. We have not built that, so selling into it would be selling exposure dressed up as insight.

And there is a measurement problem underneath it. When an assessment decides whether someone gets the job, they answer as the person who should get the job, and the data stops describing anyone real.

The underlying traits drift slowly over years, and the visible style shifts faster than the trait does. Context pulls behaviour around — a Dominance colleague in a role with no decision rights often looks markedly less dominant, and the same person given ownership snaps back within weeks. What genuinely changes with deliberate effort is not the preference for speed but the habit of checking its cost, which is a learnable behaviour rather than a personality change.

Easily, and this is why context matters more than instrument design. The items are transparent, and anyone who wants to present as more conscientious can. That's tolerable when the assessment is for the person's own development and disqualifying when there's something to win — which is one of several reasons these should never be used in hiring.

Throughline's guard is structural rather than psychometric: nothing here feeds a decision about you, so there's nothing to game.

Separate the quality from the timing explicitly, and do not ask them to lower their standard — they will resist that, correctly, and hear it as an instruction to be worse at their job. Name what the thoroughness bought, name what the delay cost, and then propose a mechanism: agreeing the required depth at the start of a piece of work rather than negotiating it at the deadline.

Lead with something good and specific, keep the criticism to one behaviour rather than a general assessment, deliver it privately, and close with explicit confidence that they can fix it. Then reconnect socially within a day or two. For a Steadiness colleague the underlying question during critical feedback is whether the relationship is still intact, and the silence afterwards does more damage than the criticism itself.

Keep the relationship visible while you deliver the substance. For an Influence colleague, warmth is not padding around the message — it is the channel the message travels on, and its absence changes what they hear. Say the difficult thing plainly, attribute it to the trait rather than to character, show the concrete cost, and give them one specific behaviour to try. Then reconnect normally within a day or two, because distance afterwards does more damage than the feedback did.

About twenty minutes per person for the three tests, four minutes to connect your assistant. Pair guidance is available as soon as two people have finished.

The honest answer on value is the first difficult message you draft with it — for most managers, inside the first fortnight. If that message isn't visibly better than the one you'd have written alone, nothing later fixes it.

No. Working style describes how someone operates, not how capable they are, and there is no defensible evidence that a particular style predicts performance in a given role. Using it to screen candidates risks filtering for people who resemble the existing team, which is both bad practice and, depending on how it correlates in your applicant pool, legally hazardous. Throughline refuses these questions outright rather than answering them carefully.

Nothing you have to do. Scoring happens as they finish, and their profile becomes available to the assistants your team has already connected.

Connecting takes about four minutes, once, per person. After that there is no dashboard to log into and no export step — you ask your assistant a question and it fetches what it needs.

EU (Frankfurt) by default for every workspace including the free tier, and London on Business and Enterprise plans. The region is fixed when the workspace is created. Model providers are the exception: depending on which assistant your team uses, profile context may be processed in the US at query time, and it is not stored there.

You do, first and fully. Your team's workspace sees the profile you confirm — and sharing is mutual: a colleague can see how to work with you only if you can see the same about them. Your results are never used to score, rank or evaluate you, and Throughline refuses to speculate beyond what was measured.

Because per-seat pricing charges you for people, not for value. Most teams test everyone once and then use the guidance every day through their assistant — and that daily use is free here. Paying per test means a team of forty that tests once a year pays a fraction of what forty seats would.

It also means nobody has to ration the tool. Add a new joiner, ask about a pair a hundred times, connect a second assistant — none of that costs anything.

Yes, any time from your original link. Retaking replaces your previous result, and your team sees that it was updated. First instincts usually produce the truest read — but the choice is always yours.

Not yet. A self-hosted deployment and customer-managed encryption keys are both scoped and neither has shipped. Enterprise plans can pin the region and run the MCP server against a dedicated database, which covers most of what people are worried about, but it is not the same thing and we will not call it that.

Full-length research instruments run to 120–300 items and take 30 to 45 minutes. Throughline uses the 50-item IPIP Big-Five Factor Markers, which take about eight minutes.

One credit is one person taking one of the three instruments. Buy them singly at $18, or in bundles that bring the price down to $10 a test at volume. Credits don't expire, and a re-test within the first month of taking a test is free — so a bad day doesn't cost you a credit.

Dominance is generally the least common in the wider workforce, at roughly one in six people, though it is heavily over-represented in senior management. Rarity carries no value judgement — it just means fewer people lead with it, and a rare style in a team is often the one being talked over rather than the one that matters most.

Yes — 20% off the Team plan when you pay annually: $17.60 a month instead of $22. There's no multi-year lock-in and no auto-escalator; if you'd rather stay monthly, that price stays available.

Their profile is deactivated and stops appearing in anyone's assistant immediately, and their assistant tokens are revoked on the next tool call. They keep their own export. You stop being billed for them on the next cycle. If they ask for full deletion, live records go within the hour and backups age out within thirty days.

There is no paid bounty programme yet. There is a monitored security address, a published disclosure policy and a commitment to acknowledge a report within one working day. We would rather answer every report properly than run a programme we are too small to staff.

Yes — 50% off the Team plan for registered charities, schools and universities. Mention it on the demo and we'll apply it without a procurement exercise.

No. Every user on every plan has unlimited MCP usage in Claude, ChatGPT or any MCP client. You pay for tests; the conversations that follow are free.

They don't, and nothing about their working life changes. There is no shadow profile, no inference from their calendar or their writing, and no partial record. A colleague who opts out simply doesn't appear, and the assistant says so rather than guessing.

You only pay for tests that are actually taken, so opting out costs you nothing.

Sharing is mutual by construction: a colleague can see how to work with you only if you can see the same about them. Nobody gets a one-way window, and managers don't get a back door.

Yes, at any time and without asking us. Every profile exports as structured JSON and readable Markdown. If you cancel, you keep the export — we are not holding your team's self-knowledge hostage to a renewal conversation.

No. Throughline will not rank colleagues by ability, advise on promotions, or speculate beyond what was measured. Ask it to and it declines — by design. The Recruitment & Hiring plan gives hiring managers working-style guidance for candidates; it never scores them.

No — and it's the right question to ask. Paste your scores into an assistant and it will produce something fluent and plausible. It will also invent the parts it doesn't know, treat a two-point gap as a settled fact, and give you advice that would read identically for a colleague with the opposite profile.

Throughline supplies measured scores for real, consenting people, the relationship between any two of them, and rules about what may not be inferred. The assistant then reasons over facts instead of filling gaps with whatever sounds right.

Three. The Big Five, using the public-domain IPIP-50 research items; a working-styles (DISC) assessment; and an Enneagram questionnaire — the last two written by Throughline and scored on scales rather than boxes. The Big Five is the most researched model in personality psychology and carries the measurement weight; the other two add the language teams actually use and the motivation underneath.

All three are read together, so where they agree the guidance gets sharper, and where they differ the profile says so. Every result shows how clear a read it is.

Because every score sits on a continuum. Throughline records how strongly each trait shows up for each person, so two colleagues who share a label still read differently — and the guidance is about that person, not the box. Our test for every piece of advice: swap in a different colleague's name and it should stop making sense.

Claude and ChatGPT out of the box, and anything that speaks MCP — including Cursor and Claude Code. Connecting takes about four minutes, once, and there's no new app to live in.

The Model Context Protocol is how an assistant reaches trusted outside data. Throughline runs an MCP server, so when you ask your assistant how to open a one-to-one with a specific colleague, it fetches that person's measured profile and the pair reading — and cites what it stood on — instead of guessing.

It's for anyone testing people who aren't on their team yet. $35 a month includes four candidate tests; beyond that you buy credits like everyone else. Recruiters use it to offer personality testing to their clients as part of the service; internal hiring managers use it to test shortlisted candidates. A dedicated dashboard for sending and tracking invites in volume is on the way.

Yes — that's the plan's main use. Send a candidate a test, and the hiring client gets a working-style profile alongside the shortlist. Extra tests are $18 each, or cheaper in bundles.