BeryX · Documentation
BeryX's technical documentation, open to evaluators and integrators alike.
The documents we maintain alongside the product code, published without rewriting: what BeryX does, how it connects to your sources and tools, and how it calculates every number.
Security, connectors, capabilities, webhooks and MCP are available in English; the other technical documents are in Portuguese.
Getting started by role
What each person on the team can do in BeryX, their first day and their weekly routine.
Getting started
Which role you have in BeryX, the app's welcome tour and where to find help inside it, with the guide for each role.
Read the documentWorkspace administrator
Connecting sources, inviting the team, approving fixes and setting up integrations: the first day and routine of whoever runs the workspace.
Read the documentMember
Connecting sources, running scans, writing rules and handling issues: what the member role allows and where to start.
Read the documentReader
Following data quality without changing anything: dashboards, reports, the assistant and when to ask an administrator for help.
Read the documentDomain owner
The owner of a data product: SLA, registered consumers, published versions, public showcase and domain approvals.
Read the documentOrganisation administrator
Single sign-on, team workspaces, access restrictions, inherited policies, quotas and the organisation's audit trail.
Read the documentBilling
The contract, the invoices and the cost breakdown by workspace, domain and team, for whoever tracks the organisation's costs.
Read the documentIntegrations
API tokens and what they connect: CI, coding agents over MCP, Excel and Power BI, webhooks and orchestrators.
Read the document
Evaluate
For RFPs, due diligence and security reviews.
Capabilities by evaluation criterion
Profiling, monitoring, matching, cleansing, metadata and orchestration, by analyst criteria — with the code and tests behind each capability, and what is still missing when it is partial.
Read the documentSecurity and compliance
Read-only access, encrypted credentials, anti-SSRF, self-service GDPR, anonymisation before the LLM, subprocessors and the status of certifications.
Read the documentMethods — how we calculate every number
Quality score, anomalies, drift, duplicates, trust score and financial impact: formulas, constants, assumptions and what is exact, estimated or sampled.
Read the documentAccessibility — WCAG 2.2 ACR
Accessibility Conformance Report (ACR based on VPAT® 2.5) for the app and the website: the 55 WCAG 2.2 level A and AA criteria, with the EN 301 549 clause, the status of each, the evidence and what is still missing.
Read the document
Connect sources and pipelines
Credentials per source, orchestrator webhooks and the quality gate in CI.
Connector guide
How to obtain credentials and configure each source — BigQuery, Snowflake, Databricks, SQL Server, Salesforce, Kafka, SharePoint and more — and the validation status of each.
Read the documentdbt Cloud and Airflow
Scans triggered by the dbt Cloud webhook, and the beryx-airflow package (operator, sensor and callbacks) to wait for the verdict in a DAG.
Read the documentQuality gate in CI
ODCS contracts exported per dataset and validated in the pipeline: a diverging schema, a score below the minimum or open critical issues fail the pipeline.
Read the document
Integrate with your stack
The API stability policy, AI agents, lineage and data catalogs.
API versioning and deprecation
What is stable in API v1, what counts as a breaking change, how and how far ahead we warn before removing something (Deprecation and Sunset headers), which SDK versions are supported and how to report a problem. Includes an English version.
Read the documentOutgoing webhooks
BeryX alerts, escalations and digests in your SIEM, PagerDuty or automation: events, body JSON Schema, HMAC-SHA256 signature with a 5-minute anti-replay window, retries and verification in Python and Node.
Read the documentMCP server
Data quality within reach of AI agents (Claude Desktop, Claude Code or any MCP client), read-only and under the workspace's RBAC.
Read the documentOpenLineage
Each scan as an OpenLineage run in Marquez, DataHub or Astro, linked to the tables it read — and your lineage inside BeryX.
Read the documentExternal catalog
Verdict, score and each rule's result published as data quality tests in OpenMetadata and as assertions in DataHub.
Read the document
Claude in your cloud account
Enterprise plan: data and token billing stay in your organisation's account.
App guide, page by page
What each page of the app is for and what you can do there — the same guide the app's Help page summarises.
Monitor
Overview
The BeryX landing page: it shows the getting-started steps until your workspace is activated, plus the dashboard for the persona you pick in the header (Engineering, Management or Executive). Use it to check the state of your data quality at a glance.
TV mode
A full-screen wallboard for the team TV, with no menu and no interaction: it shows the workspace score, critical issues, scans and the latest issues and signals, and refreshes itself every 60 seconds. It needs a signed-in session.
Datasets
The catalogue of the datasets discovered in your data sources, with each one's quality score. Use it to find a dataset, compare datasets and view the catalogue as a map, by domain or by metadata quality.
Dataset details
The page for one dataset, with its score, column profile, suggestions and open issues, organised in tabs. Use it to understand a table's quality, request a scan or an AI report and export the contract or the remediation.
Domain
The scorecard of a domain (data product): aggregate score, open issues, SLA, conformance, consumers, datasets and members. Use it to check whether the data product meets its SLA and policies and what still needs fixing.
Live scan
Shows a running scan in real time: its phases, the datasets analysed, the profiled columns and the rules passing or failing, dataset by dataset. Open it from Connections, from a dataset's details or from the sample data.
Issues
Lists the data quality issues in your datasets, with filters for severity, type, dimension and status. This is where you triage: assign an owner, understand the cause and the impact, and resolve, ignore or snooze each issue.
Incidents
Groups into a single incident the open issues that appear within the same 2-hour window in lineage-linked datasets, and points to the source dataset. Use it to tackle the cause at the source instead of treating each downstream symptom.
SLO attainment
Shows whether data arrives on time: compares, day by day, the latency promised in each dataset’s contract with the data age observed in each profile. Use it to track attainment against the objective and to explain breaches to the business.
Duplicates
Shows the groups of repeated records that scans find in each dataset, by exact or approximate match, with a sample so you can confirm them. Use it to review each group and, in the Across datasets tab, to match records from two systems.
Matching
Compares two datasets and shows which records describe the same entity, such as customers, accounts or suppliers. Use it to measure how far two systems agree; BeryX reads both sources, never writes to them, and creates no master records.
PII map
Shows where the workspace’s personal data lives: which datasets and columns hold PII, of what type, how it was detected and whether the columns are covered by active rules. Use it to review exposure with the DPO; it shows metadata only, never the values.
Act
Suggestions
Actionable recommendations generated on every scan, by heuristics and, when available, by AI. Use this page to review, approve or dismiss fixes and follow each one until a scan confirms the problem has gone.
Steward daily plan
Every day the steward gathers what needs a human decision and proposes actions ranked by risk, impact and age. Use this page to approve or reject the day's actions and set the steward's policy.
Rules
The workspace's quality rules, which the scan evaluates automatically and which open issues when they fail. Use this page to create, edit, enable or delete rules and to import rules from dbt, Great Expectations, SodaCL or ODCS.
Rule packs
A gallery of rules curated by industry, each with the reasoning behind its threshold. Use it to preview a pack against a dataset and propose the rules that make sense; they only reach the engine once accepted in Suggestions.
Activities
A board of the workspace's issues, in four status columns. Use it to triage as a team: change each issue's status, assign an owner, set due dates and see the history of changes.
Signals
Data complaints found in your team's Jira tickets, Confluence pages and chosen Slack/Teams channels, linked to the catalogue. Use this page to promote the relevant ones to issues and dismiss the noise.
Understand
AI Reports
Reports written by AI for engineering, management or executives, about one dataset or the whole workspace; with no AI available, they are produced from an offline template. Use this page to request reports and open each one to read, export or share it.
Routines
This page shows, step by step, how an AI report is made: where the data comes from, what gets checked and who writes the text. Use it to understand a specific report, see where it failed and adjust scheduled reports.
Knowledge
The team's memory about the workspace's data: summaries of reports and diagnoses, how problems were resolved and the decisions recorded. Use it to search what is already known, add notes and see how sources, datasets and quality connect.
Glossary
The workspace's business terms, such as "Net revenue", and the columns that implement them. An approved term, with its link to the column accepted, goes into that column's ODCS contract as businessName and inherits the quality of its linked columns.
Reference data
The workspace's code lists, such as countries, currencies or order statuses, with versions that never change once created and mappings from other systems' codes. “Reference list” rules validate columns against them, and coherence shows where datasets disagree.
Assistant
Ask questions in plain language about your workspace's datasets, issues, costs and reports and get answers from that data. Nothing changes without your approval: the rules and issues the assistant proposes are only created if you approve them.
Connect your agent
Connect your coding agent to BeryX so it checks each table's quality, contract and PII columns before touching it. The page gives you ready-to-paste configuration for each tool, read-only.
Compliance
Gathers the AI Act Article 10 evidence for your AI systems: you register each system, assign its training, validation and testing datasets, and see what BeryX measures and what only you can declare.
Governance
Shows whether the governance programme is working: catalog coverage, issue resolution, SLAs, adoption, quality by dimension, incidents and usage, each KPI against its goal. Use it to follow the programme and report on it.
Connect Excel / Power BI
Connect Power Query, in Excel or Power BI, to stable tables of the workspace's scores, issues, SLO, costs and activity, with a read-only token. Use it to run your own analyses without asking the data team for anything.
Administer
Connections
This is where you connect BeryX to your data sources, always read-only, and decide when each one is analysed. Use this page to add a source, test the connection, discover tables or request a scan.
Team & Access
This is where you manage who joins the workspace and what each person can see: users, teams, access by domain, API tokens, two-factor authentication, AI and plan. Only administrators make changes on this page, except in the “Organisation” tab, which belongs to people with an organisation role.
Notifications
Send BeryX alerts to where the team already works: Slack, Microsoft Teams, email or a webhook. This is also where you set up on-call rotations, Jira or Linear tickets and the git repository for approved fixes.
Presence
Decide what BeryX posts in your team's tools, where, and with what content. Today only Slack receives posts and has a bot; Teams, GitHub, Confluence and Jira appear with what BeryX already does with them. The page is for workspace admins only.
Audit log
Shows who changed what in the workspace: members, plan, connections, tokens, channels and BeryxTech support access, with sensitive events highlighted. Only workspace admins can see it.
BeryxTech Console
For the BeryxTech team only; customer accounts do not see it. It brings together customers, organisations, platform metrics and support access, using metadata only: a customer's data can only be read with access approved by their admin.
Help & integrations
Brings together what BeryX can read, how we calculate each number, how to connect AI agents, Excel and Power BI, a short guide to every page and where to ask the BeryxTech support team for help.
Account and access
Sign in and activate your account
This is where you sign in, create a free account or recover access. It is also where you confirm your email and set your password from the links we send you, and complete two-step verification.
Account security
Protect your account with two-factor authentication: besides your password, signing in asks for a code from your authenticator app. Here you turn it on, manage your recovery codes and, if the policy allows, turn it off. You also see and end your open sessions.
Connect your Slack account
Connect your Slack account to your BeryX account to ask the BeryX bot questions in Slack. The same page connects your Microsoft 365 account to the BeryX agent in Microsoft 365 Copilot. It opens from the link the bot or the agent sends you and also lists the accounts you have already connected.
Public pages
Verified data quality
A public page, open without signing in to anyone with the link, with the quality summary of a dataset its owner chose to share: score, verdict, dimensions, failing checks and data contract. It never shows samples or personal data.
Data product showcase
A public page, open without signing in to anyone with the link, with what a consumer needs to decide whether to use a data product: the declared SLA and this month's attainment, readiness per use case, and each dataset's verdict and score. It never shows samples, columns or personal data.
Data contract proposal
A no-sign-in page, opened from the link you were sent, where you review a dataset's proposed contract section by section as the owner of the source system. Once every section is accepted, you accept the contract and BeryX starts checking it.
Subprocessors
A public page, no sign-in needed, listing the companies BeryxTech entrusts with part of BeryX's data processing: each one's purpose, data, region, legal basis and DPA, plus the changes announced in advance.
DORA and NIS2
A public page, no sign-in needed, with short answers for RFPs from financial entities (DORA) and essential or important entities (NIS2). It sums up what BeryxTech gives regulated customers; it is not a compliance statement or a contract.