Documentation

Anchord is the read-only identity layer you call before writing to any external system. Resolve records to canonical identities, guard every write, and escalate ambiguity to humans instead of guessing.

What Anchord does

Resolves identity — matches records from any system to a canonical AnchorID using rules-based scoring
Guards writes — evaluates a proposed write and returns allowed or blocked without side effects
Escalates ambiguity — when multiple candidates score too close, returns needs_review so a human can decide
Detects duplicates — clusters near-duplicate entities and surfaces them for merge with confidence scores
Never writes back — read-only by design; ingests data from HubSpot, Salesforce, Stripe, and custom sources but never modifies them

Start here

I'm building an agent

Set up the MCP server or call the REST API. Resolve identity and guard writes in your agent's workflow.

Agent Quickstart →

I'm setting up Anchord for my team

Connect CRMs, explore your data, manage identity quality. Load demo data to see the product immediately.

Full Quickstart →

I want to explore the API

Browse all endpoints, see request/response schemas, and try calls from your browser.

API Reference →

Guides

Agent Quickstart

For agents

MCP install for Cursor and Claude Desktop, working examples (resolve, get entity, guard write), 11 MCP tools, needs_review handling, and a smoke test. Single page — everything an agent or builder needs.

Quickstart

End-to-end onboarding for humans: sign up, get an API key, ingest records, resolve identities, guard writes, connect Salesforce/HubSpot/Stripe, and merge duplicates.

Integrations Guide

Connect Salesforce, HubSpot, and Stripe via OAuth. Run syncs, configure backfill windows, push data via the ingest API, manage custom sources, and troubleshoot.

Scenarios

Concrete workflows for specific jobs-to-be-done: before updating HubSpot, creating Salesforce records, syncing Stripe data, handling ambiguity, and preventing duplicates.

Recipes

Three copy-paste curl flows for common agent patterns: resolve-before-write (ingest → resolve → guard), guard-before-CRM-update, and needs-review escalation.

API Reference (Swagger)

External →

Full interactive OpenAPI documentation. Browse all endpoints, see request/response schemas, and try calls from your browser.

API quick reference

Endpoint	Description
GET /api/ping	Health check / verify API key
POST /api/v1/ingest/batch	Ingest up to 100 records from any source
POST /api/v1/resolve/company	Resolve company → canonical AnchorID
POST /api/v1/resolve/person	Resolve person → canonical AnchorID
POST /api/v1/guard/write	Evaluate a proposed write (allowed / blocked)
POST /api/v1/guard/write:batch	Batch guard check (max 200)
GET /api/v1/entities/{id}/export	Export golden record with provenance
GET /api/v1/entities/{id}/targets	Per-system target IDs for an entity
POST /api/v1/sources	Register a custom source system
POST /api/v1/entities/{id}/merge	Merge duplicate entities (dry-run supported)

Full interactive docs at api.anchord.ai/docs