Skip to content
Cupigraph

Product

A matching engine that keeps its reasoning visible.

Cupigraph stores what you know as entities and relationships, applies the criteria you set, and returns suggestions with the paths that support them. Your reviewers take it from there.

Graph model

Everything you match on, and how it connects.

Records you already keep become entities. The facts between them become relationships, each with a source you can point to.

Entities

The things you match and the things that connect them.

  • People
  • Roles
  • Organizations
  • Requests
  • Skills
  • Places
  • Programs
  • Projects

Relationships

Typed, directed edges that say how two entities relate.

  • requires
  • practices
  • offers
  • located in
  • member of
  • contributed to
  • invests at

Provenance

Every edge records where it came from and what it may be used for.

  • source
  • date
  • evidence
  • consent scope
  • visibility
cupigraph · relationship
Example · fictional data

Edge

Candidate A-104 —practices→ PLC programming

source
Profile import · Sep 12, 2026
evidence
6 years, stated by the candidate
consent
Matching for open roles
visible to
Hiring team for this role
status
Active

Criteria

You define what a good fit means.

Pick the paths that count, mark the ones that are required, and weigh the rest. Exclude any field that should never play a part. Cupigraph applies exactly that definition, and nothing else.

  • Criteria are readable by the people they affect.
  • Changes are versioned, so you can see which definition produced a suggestion.
  • Different teams can keep different criteria for different decisions.
cupigraph · criteria
Example · fictional data

Required

Role —requires→ Skill ←practices— Person

Preferred, with weights

  • Role —team uses→ Tool ←has used— Person

    0.30
  • Role —based in→ Region ←open to— Person

    0.20
  • Role —works on→ Project type ←has shipped— Person

    0.15

Never traversed

  • date of birth
  • photo
  • marital status
  • home address
cupigraph · explanation
Example · fictional data

Controls Engineer ↔ Candidate A-104

for review

Supporting paths

  • Role —requires→ PLC programming ←practices— Candidate

    updated Sep 2026

  • Role —team uses→ ROS 2 ←built with— Robotics project ←contributed to— Candidate

    updated Jun 2026

Gaps

Role —requires→ Work authorization · not on file yet

Conflicts

Start date: the role needs January, the candidate is available from March.

Each piece of evidence shows when it was last updated.

Explanations

Every suggestion shows its working.

An explanation is built from the graph itself, so each line points to a real relationship a reviewer can check.

  1. 1

    Supporting paths

    The routes through the graph that connect both sides, strongest first.

  2. 2

    Gaps

    Requirements with no path yet, so the right question gets asked.

  3. 3

    Conflicts

    Places where the two sides disagree, shown rather than averaged away.

  4. 4

    Freshness

    When each piece of evidence was last updated, so stale facts stand out.

Review and decisions

Cupigraph never accepts, rejects, or contacts anyone on its own.

It prepares the evidence. A person reads it and decides.

Suggestions, not verdicts

Each match arrives marked for review, with its paths, gaps, and conflicts. Nothing moves forward until a person acts on it.

Decisions with reasons

Shortlist, pass, or ask for more. Every decision records who made it, when, and the note they left.

Corrections at the source

If a reviewer marks a relationship as wrong or out of date, it is fixed where it came from, not hidden in one result.

Consent and privacy

Data is used only for what it was shared for.

Consent lives on the relationship itself, so the rules travel with the data instead of sitting in a separate policy.

  • Each relationship stores the consent scope it was shared under.
  • When consent is withdrawn, the relationship stops being used in matching.
  • Fields you mark as excluded are never traversed, for any criteria.
  • Visibility rules decide which reviewers can see which evidence.
  • Your graph is used for your matching, not pooled with anyone else's.

API and webhooks

Fits into the tools you already use.

Send records in, ask for traces between two entities, and receive suggestions in your own review tools. Webhooks tell you when a reviewer has made a decision.

Notice the decision field: it stays empty until a person fills it.

api · /v1/traces
Example · fictional data
POST /v1/traces
Content-Type: application/json

{
  "from": "role_8812",
  "to": "person_a104",
  "criteria": "controls-engineer"
}

See your own criteria as a graph.

Bring an example of what you match today. We'll walk through how it would look in Cupigraph, path by path.