Housing assessment without the black box.
HEAT, the Housing Emergency Access Tool, is a structured housing triage assessment for coordinated entry: it documents housing needs, computes chronic homelessness status, and identifies interventions to discuss, without assigning a vulnerability score or deciding who gets housing.
Every determination explains itself.
Public beta in development. Not for production decision making.Try the assessment preview Explore a shadow pilot
New here? See four sample results first, no questions asked.
- 23 questions everyone is asked
- 36 hard cap, so no one is ever asked more
- 26 to 30 questions the four published samples were asked
- 4 support planning bands, named and never scored
Every number above is read from instrument 0.25.0-pilot or computed by running the published samples through the real engines, never typed by hand.
What HEAT produces
One made-up person, put through the same questions and the same engines a real assessment uses. This is the output, not a picture of it.
An adult, sleeping outside as of last night, in month 13 of this period without housing.
- Immediate action
- No immediate emergency identified
- HUD homeless definition
- Category 1: literally homeless
- Chronic homelessness
- Meets, based on entered information
- Documentation still to collect
- 2 documents, listed with the counting rules behind them
- Support planning level
- Intensive stabilization planning
- Interventions to discuss
- Permanent Supportive Housing (PSH): discuss firstRapid Re-Housing (RRH): considerEmergency shelter: discuss first
On the real page every row here opens on the rule that produced it, the answers it read, and what is still missing. Nothing on a result is unexplained.
Assess. Explain. Recommend. Support.
Four outputs, never one score. Every assessment ends with what to pursue, what documentation is still needed, what support to expect after housing, and what to do next.
01Immediate action
A neutral safety screen that routes emergencies to live help first. Workflow only, never allocation.
In the sample above: No immediate emergency identified.
02Chronicity, computed
HUD chronic homelessness status under 24 CFR 578.3, with the arithmetic shown and a checklist of the documentation still needed.
In the sample above: Meets, based on entered information, with 2 documents left to collect.
03Support planning level
Categories, not scores. Missing or declined answers are reported as missing, never counted against anyone.
In the sample above: Intensive stabilization planning, one of 4 named bands.
04Intervention discussion
Options to discuss, each with reasons, open eligibility questions, and cautions. Decision support, never automated placement.
In the sample above: Permanent Supportive Housing (PSH) to discuss first.
HEAT assesses. Your community decides.
The boundary is the design, not a disclaimer. Everything on the left is measurement, published and reproducible. Everything on the right is policy, and it is yours.
HEAT
- Establishes facts: statutory eligibility, housing history, and what support would help.
- Computes chronic homelessness status under 24 CFR 578.3, with the arithmetic shown.
- Identifies the documentation still needed to prove it.
- Assesses support planning needs as named categories.
- Identifies interventions to discuss, with reasons, open eligibility questions, and cautions.
- Explains every determination, rule by rule, in words a reader can check.
Your CoC
- Sets prioritization policy, in its own adopted written standards.
- Determines queue position. No score exists inside HEAT to compete with it.
- Allocates resources, including who is offered the next opening.
- Makes referral decisions, through its case conferencing process.
HEAT is designed to support housing prioritization by producing standardized, transparent, clinically meaningful findings. It does not determine queue position or allocate housing resources; those decisions remain with your written coordinated entry policies and case conferencing process. HEAT never overrides local policy or HUD requirements, and it never ranks people.
Built to be checked
- Deterministic. Identical answers always produce identical results, byte for byte.
- Transparent. Every result prints its full rule-by-rule reasoning, and every question is public with its evidence and known risks. Nothing on a result is unexplained.
- Private. In public mode, answers never leave your device. Nothing about you is stored unless you choose it.
- No black-box scoring. No numeric priority score exists anywhere in the tool. When measurement is appropriate HEAT measures, shows its work, and reports its confidence. When policy decides, your CoC decides, in your CoC's own words.
Also true, and published rather than compressed away: no AI in any determination, so every conclusion comes from published rules a person can read; medicine measures constantly, and what fails is opaque, unvalidated ranking; the only thing counted here is a bare total of page visits and completed assessments, with no identifiers; and every question, every rule that fires, every threshold, and every mistake is public on this site, though the software itself is proprietary. All of it, in full, on the governance page.
Do not replace your assessment. Test HEAT against it.
Switching assessments is a governance decision. Piloting HEAT is not. In a shadow pilot, HEAT runs after your current assessment, asks everyone the same 23 questions with a few more depending on the answers, and affects no referrals. Your case conference compares the two outputs side by side: which one explains itself, which one arrives with the chronic homelessness documentation checklist already computed, which one your staff trust more. Every completed HEAT arrives at case conference as a ready briefing: the facts, the open questions, and the documentation status, already on one page. The comparison data is yours, and it doubles as validation evidence.
Standing challenge: find a scenario where you believe HEAT reaches the wrong determination. Send it through the Feedback button, with your name or handle if you want credit. Every accepted case joins the permanent regression suite before any fix is written.
Are you taking a HEAT assessment?
- Your answers stay on your device. Nothing is saved and nothing is sent anywhere. Closing the page erases everything.
- You can decline any question. Every one of them takes "I do not know" and "I prefer not to answer". Nothing you decline is counted against you, and it is never treated as a "no".
- There is nothing to pass. No score, no ranking, and no test. HEAT does not decide who gets housing and does not hold a place in any line; people decide that, using their community's own written rules.
- You get something at the end. A page you can print or hand to a housing help office, so you do not have to tell the whole story again, plus what to do next and what is still missing.
See what that page looks like, with made-up people, no questions asked.
Evidence and accountability
Nothing above has to be taken on trust. Each of these is a published page, and each one is regenerated or gated by the build.
- Questions. Read every question we ask, in the order they are asked, before you answer any of them.
- Evidence register. The evidence behind each one, what it changes, and its known risks. A question with no entry fails the build.
- Governance. How decisions are governed, where the boundaries are, and how to disagree with a result.
- Changelog. What we changed and why, mistakes included.
- Public register. Proposed changes, open for comment before anything is decided.
Where this is going: public beta now, then shadow mode alongside a CoC's existing assessment, then HMIS-assisted confirmation instead of re-asking what is already known, then a full case-conference deployment. Deterministic engines are live today for immediate action, HUD homeless category, chronic homelessness, support planning, soft supports and reassessment guidance. Next: cognitive testing of question wording, and pilot recruitment. What is live, and what changed, is in the changelog.