Credda, Inc.

Portable, verifiable trust infrastructure.
<sub>A 0 to 100 reliability record built only from outcomes somebody else confirmed,
owned by the person it describes, and scored by a formula anyone can read.


The first three are not decoration. They are read out of production every hour,
so they go stale loudly instead of quietly.</sub>

credda.io  ·  API reference  ·  the scoring formula  ·  help center

Your track record is stuck wherever you earned it. It lives inside one platform, in a shape that platform controls, and it does not come with you when you leave. Start somewhere new and you start at zero.

Credda gets it out. Platforms report outcome events to an append-only ledger: a contract delivered on time, an invoice a client accepted, a marketplace job finished, a shift a supervisor signed off, a pull request someone else merged. A pure function turns that ledger into a reliability record from 0 to 100. The record belongs to the person it describes. You carry it between platforms, present it as a signed credential, and decide how much of it anyone gets to see.

It is built for people whose work other people depend on: freelancers, contractors, gig and marketplace workers, shift workers, engineers, small businesses, and increasingly agents acting on someone's behalf.

What Credda is not

Credda does not rate, judge, screen, or decide anything about a person. It verifies evidence and records it. What to do with that evidence is the reader's call. There is no "should I hire, lend to, or trust this person" verdict anywhere in the product, from a human or from a model. We left it out on purpose, and we are not adding it later.

Published, not proprietary. Every number below is served live by GET /api/v1/scoring/model, which reports the formula version the engine is actually running. Read it rather than trusting this table.

The five weights sum to 1.00. Verified Professional Grounding is binary, capped, employer-agnostic and duration-blind: it asks whether anyone has verified that you have held a professional role, never where, never for how long, and never how prestigious. It cannot rescue a failing record, only lift a non-failing one, and by at most its own weight.

Verification Depth asks how much of your record someone else confirmed, not how many logos you collected. A fully verified record on a single platform earns full marks.

A record starts with no verified evidence and is earned upward. Verified outcomes are what unlock the top of the scale, so a record built only on unverified activity tops out at Proven. One serious breach drops it sharply, because a high score has to be earned and we do not cushion the way down.

At Risk is reserved for evidence, not for the absence of it. A record with no third-party-verified outcomes is floored, so an unverified adverse outcome cannot on its own put anyone in the worst band. Only a confirmed one can. That floor is a switch, and its state is published rather than asserted, on the provisional.unverifiedFloor block of the scoring model endpoint. Read it there rather than here.

Billing is live and checkout is self-serve. It buys API access and a short list of conveniences. It does not touch anyone's record, and that is enforced in one file rather than promised in a paragraph.

Credda is larger than the marketing site suggests. The platform API alone is 394 route handlers across 40 modules, and the scoring service is a separate codebase again.

Three public endpoints answer that, without an account and without asking us. They are derived from the running configuration, not from a page someone remembered to edit.

If any of those disagrees with something written here, the endpoint is right and this page is stale. Tell us.

These are the lines we do not cross. We publish them because a trust record is only worth as much as the guarantees behind it.

• Deterministic and bias-free. No human and no model decides a score. There is no manual override, no adjustment dial, no adjudicated appeal. A dispute resolves by the same rules as everything else.
• A score cannot be bought. A paid plan governs API access and the ten conveniences above. No tier and no amount of money moves anyone's record.
• Worker-owned and portable. You hold the credential, you choose what to disclose, and you can revoke it.
• Never a verdict on a person. We explain evidence. We do not recommend decisions about human beings.
• A third-party witness is required. isVerified is never granted to a party vouching for itself. A bare payment is not trust. A self-confirmed job is not trust.
• An employer cannot push a bad mark onto your portable record. When a business records a missed deadline or a breach against someone, that stays on that business's own internal record. You see it, you are told about it, and you can contest it, but it does not follow you to the next employer. Carrying employer-furnished negatives between companies is a different product with different obligations, and it is switched off until we can meet them.

Everything is contract-first, described in OpenAPI 3.1 and rendered at /docs.

All four client libraries are MIT licensed, and the three npm packages are published with build provenance.

npm i @credda/js

go get github.com/Credda-io/credda-go

Credda is early. We do not publish user counts, customer names, revenue, or funding, because we would rather say nothing than say something we cannot stand behind. The scoring service, web app, credential fabric, and developer platform are all live at the links above, and self-serve checkout is open. Whether you can buy a given plan today is answered by /subscription/pricing, which reads the running billing configuration, so it cannot tell you a plan is available when it is not.

Security and compliance readiness work (GLBA Safeguards Rule, SOC 2) is in progress. Credda is not certified or attested against either framework, and we will not describe it as such until an auditor says so.

Found a vulnerability? Read the security policy, then email security@credda.io with the subject "Security disclosure". The same contact is published under RFC 9116 at credda.io/.well-known/security.txt, so you can confirm it against the service rather than against this page. Please do not open a public issue for a security report.

Anything that is not a vulnerability: where to get help says which channel handles what. Questions about your own record belong in the help center, never in a public issue.

_{Open about how it works. Closed about your data.}