What's new at Credda

Shift work in the organization console: record a shift, stage the workers who covered it, hand each of them a claim page, and give a supervisor the role that confirms it.

Staff can verify an organization, with the corroboration signals behind that decision shown beside it rather than left to judgment.

People can describe confirmed work in their own words, marked as their own account of it, so a record can carry detail without that detail being mistaken for something a third party attested.

A witness who cannot confirm something can correct it instead of going silent, and the correction sits beside the record rather than over it. A decline is recorded as a decline and never counted as a confirmation, and one witness gets one ask.

Plain language on the pages a confirmer sees, written for people who do not work in software, and raw error codes no longer reach them.

Accessibility across the core loop and the stranger-facing confirmation paths: every navigation is announced, focus moves with it, sign-in errors are spoken as well as shown, form fields are named, and the smallest controls now meet the minimum target size.

A record with nothing confirmed yet reads as nothing confirmed yet, rather than as zero percent, and a genuine small measurement no longer rounds away to nothing.

An imported work history is no longer presented as a confirmation, a lapsed credential stops looking current, and an organization record can no longer be described as an average of the people in it.

Collect Confirmations: an organization can work through its client book and ask each client to confirm the work that was actually done.

A guided, resumable walkthrough for a new account, and a getting-started spine for a new organization, so neither starts on an empty screen with no next step.

Public person and company pages lead with a strength headline and a clear next action, show how densely a record has been confirmed and by whom, and a company record can be published at its own branded address.

Verify a professional license against the national register from the web app, and embed a profile card or a milestone in your own site.

Public profiles are properly indexable: structured data, per-page metadata, a sitemap and robots rules. A profile nobody chose to publish stays out of the index.

The six score bands were renamed to match the names the scoring API serves, so the same record cannot be described two different ways depending on which surface you read.

A score component with nothing to measure yet says so, instead of showing a zero that reads as a failing mark.

Search split in two: a command palette for your own account, and a separate page for looking somebody else up.

Verify my career: ask every past employer, client and school on your record to confirm their part of it, in one batch rather than one message at a time.

A shareable, verifiable company record for organization owners, alongside adverse-outcome reporting and recurring requirements such as annual training and renewals.

Credda installs as an app, and a use-cases section says plainly where a verified record is actually used.

The pricing page was split into three deliberate products, and the team score board no longer ranks colleagues against one another.

The public profile no longer prints a rank verdict as its headline, education dates appear, and a school still in progress reads as present rather than blank.

Scoring v5.3: new accounts start at 20, in a new Provisional band, instead of near a neutral baseline. A high score is earned across roughly six verified outcomes; verified failures are never cushioned or averaged away, and verification depth now rewards the share of your record that is independently confirmed, never how many platforms it spans.

The developer platform is live at api.credda.io: docs, an interactive API reference, an OpenAPI 3.1 spec, and a self-serve console for API keys, usage and webhooks.

Badges and achievements: deterministic, display-only milestones (fulfilled-commitment volume, verified platforms, flawless streaks) on public profiles. Never a scoring input.

Global search across people and organizations, plus organization trust scores, team score boards and org-level commitments.

Published to npm: @credda/js (TypeScript SDK), @credda/cli (command line, including offline credential verification), and @credda/mcp-server so AI agents can check trust mid-reasoning.

Scoring engine v3.1: a commitment left undelivered past its deadline now counts as a breach that lowers your completion rate, instead of a masked "completion". Delivered-late still counts as fulfilled but affects your on-time rate.

Scores respond to real activity sooner; the confidence baseline no longer holds new accounts near a neutral 50 for as long.

Higher-contrast text, especially in dark mode, for better readability across the app.

AI Trust Coach is live: plain-language, score-aware coaching on your dashboard that names your biggest opportunity and the projected gain from your next on-time commitment. Advisory only; AI never decides a score.

Disputes are now resolved by a fully deterministic engine, from the tamper-proof commitment record alone, with no human reviewer picking the outcome. This removes human bias from the trust score entirely.

Fixed profile pages that could show "profile not found" for valid accounts: profile URLs are now case-insensitive, plus a database schema backfill for the commitment lifecycle.

Public verify endpoint and rotatable share tokens: any site can embed a live trust badge with no API key.

Rebuilt the embeddable trust badge: Shadow-DOM isolated, light and dark, and badge, compact and inline variants.

Scoring engine v3.0: completion is now weighted by transaction value, platform trust is blended across every platform rather than winner-take-all, and a new momentum factor reflects an improving or declining trend.

Typed JS and React SDK (@credda/js): resolveToken, getScore, getScoreExplain, getScoreHistory and getPlatforms.

Organizations: teams with roles (owner, admin, member) and a public organization profile.

Every profile gets a permanent, privacy-aware timeline of verified milestones.

Backend refactored into clear modules with an event-driven trust engine underneath.

Provider-abstracted AI framework (Claude), plug-and-play and advisory-only; AI never makes a trust decision.

Public profiles with reliability score, connected platforms and commitment history.

Platform connections via GitHub and GitLab OAuth, with Google sign-in.

Interactive API documentation with expandable endpoints and copy-ready examples.

Score explanation endpoint returns per-factor descriptions and a confidence breakdown.