AI-ready MCP endpoint
A single endpoint built for AI agents, on the open Model Context Protocol — connect your own agents or an off-the-shelf assistant and work with your governed data through one secure, standard connection.
Born for the Agentic AI Era
The One was built for AI from day one. Your agents — and the assistant built into the product — work over the very same governed data your applications do, with the signed-in user’s own permissions, every request re-checked and sensitive values masked. AI is a first-class consumer here, never a back door.
A single endpoint built for AI agents, on the open Model Context Protocol — connect your own agents or an off-the-shelf assistant and work with your governed data through one secure, standard connection.
A first-party assistant lives right inside the product — ask it to find a record, explain a match, surface duplicates, or summarize an entity, in plain language and in the context of the page you are on.
Whether the request arrives over the REST API or the MCP endpoint, the same four governance layers run on the signed-in user’s identity. An agent can never see — or do — anything its user couldn’t.
The assistant grounds every answer in real platform data and points to the source behind each claim; masked values stay masked, and it will say “I don’t know” rather than invent one.
Replies render as live, interactive cards — a record’s 360° profile, a data-quality gauge, a side-by-side match comparison, a merge timeline — not walls of text to copy and paste.
Before the assistant changes anything — merging records, editing a value — it shows exactly what will happen and waits for a person to approve. How much it may do on its own is yours to configure; its authority never exceeds the user’s.
Point the assistant at a record to surface its potential duplicates, walk a merge, or explain why two records scored as a match — the same engine your stewards use, at conversational speed.
Matching & unification
Your customers, suppliers, and products live in a dozen systems, described a dozen ways. The One unifies them into a single, trusted golden record — and shows its work every time.
An intelligent matching engine resolves duplicates and near-duplicates by weighing the real evidence in the data — far beyond exact, naïve comparison — and finds candidates fast, even across millions of records.
Every match shows its work — which attributes agreed, how much each one counted, and the resulting confidence. Auditable by construction, never a black box.
Configurable survivorship rules pick the best value for each attribute across sources — source priority, most recent, most frequent, or keep-all — producing one authoritative record per entity.
Every merge is lossless and fully reversible — relationships and history re-point automatically — with complete lineage back to each contributing source record.
Dry-run any match to see the candidate pairs and the proposed golden record with nothing written — stewards decide with the full picture in front of them.
New records match automatically the moment they arrive; ambiguous pairs route to a data steward to confirm, reject, or snooze in a review queue.
Matching, calibrated
Stewards calibrate matching with evidence, not guesswork — and roll changes out safely, never disturbing what is live.
Stewards label true and false matches; the platform measures precision, recall, and F1 against that ground truth — so match quality is a number, not a feeling.
Let the platform find the auto-match and review thresholds that hit your precision-and-recall target — no manual trial and error.
Test how a rule or threshold change would have scored your labeled examples before you ever apply it to live data.
Run two matching strategies side by side per domain — stage, compare, and promote a new one without ever disturbing what is live.
Common values like Smith or John are automatically down-weighted, so the matcher never over-trusts the things everyone shares.
Data quality
Before any record is matched or served, it is parsed, standardized, validated against your rules, and scored for trust.
Names, emails, phone numbers, dates, and global postal addresses (60+ countries) are parsed and normalized into one consistent, comparable form.
“Sounds-like” matching catches spelling and transliteration variants — Catherine and Katharine resolve to the same person.
Format, range, length, lookup, uniqueness, and cross-field rules enforce your data standards — each with an error, warning, or info severity.
Every record carries a trust score across completeness, accuracy, consistency, timeliness, and uniqueness, weighted to your priorities.
Track data health over time and across domains, so a clean-up initiative has a single number to move — and a way to prove progress.
Data modeling & 360°
Define your own entities, attributes, relationships, and hierarchies — and see each one as a complete, connected 360° view, at any point in time.
Master customers, suppliers, products, locations, or any domain — entity types, attributes, and relationships are configured, never coded.
Connect entities into relationships and roll them up into hierarchies for a full picture of how your data relates.
One request returns the complete entity — its golden values, every contributing source value, its relationships, crosswalks, and merge history.
Record interactions — calls, visits, transactions — against entities for an operational, time-ordered history on every record.
Hold many emails, phones, or addresses per record, each tagged by purpose (home / work / mobile) with a preferred value.
Every golden record keeps its crosswalk back to each source system's own identifier, so nothing loses its origin.
Every record is fully versioned — see any entity exactly as it stood on any date.
Reference data
Codes, lookups, and classifications are first-class, governed data — versioned, hierarchical, and mapped from every source.
Govern canonical lists — countries, currencies, statuses, classifications — with labels, ordering, translations, and effective dates.
Map each source system's values to your canonical set, so “US”, “USA”, and “United States” all resolve to one.
Organize reference data into multi-level hierarchies for roll-ups, drill-downs, and consistent classification.
Reference sets are versioned over time — schedule future changes by effective date — and common international standards ship built in.
Governance & stewardship
The right data to the right person, enforced on every request — through your applications or your AI. A value the caller isn't entitled to never leaves the platform.
Isolation enforced at the data layer — below the application, below the AI, below anything a request can talk its way around.
A managed permission catalog; each caller acts with their own identity and never more authority than the person who signed in.
“This steward, but only customers in her region.” Access is evaluated on who, what, the action, and the context — on every read. Deny always wins.
Sensitive fields are masked before any response leaves the platform — full, partial, hashed, or redacted, by configuration and by role.
Connect your existing identity provider; people sign in once with no passwords stored, and inherit the right roles by group — automatically granted and just as cleanly revoked.
Issue instantly-revocable tokens for system-to-system integrations, kept entirely separate from people’s identities.
Compliance & lineage
Who changed what, who saw what, and exactly when — in a complete, tamper-evident trail you can always answer to.
Every change and every access is recorded in a write-once, tamper-evident trail — the platform itself won't let it be rewritten.
Every value on a golden record traces back to the source record and the merge that produced it.
Built-in consent records and handling for access, erasure, and portability requests under GDPR, CCPA, and similar regimes.
Capture point-in-time snapshots of your rules and configuration, and compare exactly what changed between them.
Real-time delivery
The same trusted, governed data, delivered to your applications in real time — and pushed to your downstream systems the moment it changes.
A standards-based REST API delivers each golden record and its full 360° view to your applications in real time — every call governed on the caller’s own identity, with field masking applied before any response leaves.
Outbound event notifications keep downstream systems in sync the moment your data changes — no polling, no nightly batch.
Enterprise platform
Cloud-native, multidomain, and multi-tenant — enterprise scale, with industry-ready starting points and without the enterprise lock-in.
Master every domain — customers, suppliers, products, locations — in one platform, with hard isolation between tenants.
High-volume bulk load, update, and merge run as managed jobs, matching automatically as data arrives, with progress you can watch.
Fast fuzzy and full-text search finds and resolves records across all your data.
Pre-built configurations for banking and healthcare — entity models, matching, rules, and roles — get you live in days, not months.
Elastic, cloud-native scale on an entirely open foundation — no proprietary lock-in.
The payoff
Validated, matched, governed data to any agent your teams build — without a single agent ever touching a system of record.