Use case
Prevent Duplicate Records Across Systems
Every system has its own duplicate detection, but none of them see records in other systems. When an agent syncs a Stripe customer into Salesforce, neither system knows the record already exists in HubSpot. Anchord prevents cross-system duplicates by matching every record to a canonical AnchorID before any write happens.
The cross-system duplicate problem
HubSpot knows about HubSpot contacts. Salesforce knows about Salesforce accounts. Stripe knows about Stripe customers. But when your agent or pipeline moves data between these systems, none of them can tell you the record already exists somewhere else.
The result: the same company appears as three separate records across three systems, each with slightly different data. Reports are wrong. Billing is fragmented. Customer communication is inconsistent. And every new sync run risks creating more duplicates.
How Anchord prevents cross-system duplicates
Every record ingested into Anchord — from any source — is matched to a canonical AnchorID. Before writing to any external system, the agent or pipeline calls resolve. If the record matches an existing entity, the agent updates the existing record instead of creating a new one.
Ingest from all sources
Connect Salesforce, HubSpot, and Stripe via OAuth. Push records from custom systems via the API. Anchord sees all records in one place.
Resolve before writing
Call POST /api/v1/resolve/company
or /resolve/person.
If the record matches an existing entity, you get the entity ID back. Update the existing
record instead of creating a new one.
Handle ambiguity safely
When signals conflict, Anchord returns needs_review
instead of guessing. A human resolves the ambiguity in the console. No wrong merges,
no wrong updates.
Rules-based scoring
Anchord uses rules-based scoring signals to match records. Domain matches score highest. Name + location, email, provider-specific IDs, and alternate identifiers accumulated over time are all considered. Each signal has a defined weight — there's no opaque ML model. Entities build a golden record as more sources link to them, improving matching over time.
When two candidates both score above the confidence threshold and fall within a narrow
margin of each other, Anchord flags the case as ambiguous rather than picking a winner.
This is the needs_review
mechanism — the system admits uncertainty instead of hiding it.
Duplicate detection and merge
Beyond pre-write resolution, Anchord clusters near-duplicate entities across all connected sources. The duplicate queue in the console shows clusters ranked by confidence, with a recommendation for which entity should survive the merge.
Merge with a dry-run preview first — see exactly which source records will be repointed, which identifiers will be combined, and whether any collisions exist. Then confirm the merge via the console or the API. Merged entities are tombstoned with a redirect, so existing references continue to resolve correctly.
Stop duplicates at the source
Resolve identity before every write. Catch duplicates that no single system can see on its own.