# Anchord

> Read-only identity resolution and write-guard layer for AI agents.

## What Anchord does

Anchord sits between an agent and the external system it is about to write to.
It resolves records to canonical identities, evaluates proposed writes, and
returns a decision — never writes back.

## Primary workflow: resolve before write

1. Ingest — push records via POST /api/v1/ingest/batch, or connect HubSpot, Salesforce, Stripe via OAuth.
2. Resolve — POST /api/v1/resolve/company (or /person). Returns: resolved, not_found, or needs_review.
3. Guard — POST /api/v1/guard/write. Returns: allowed or blocked.

If resolve returns needs_review, the agent stops and a human resolves
the ambiguity in the Anchord console. The agent retries after resolution.

## Core use cases

- Before updating a HubSpot company or contact
- Before creating a Salesforce account or contact
- Before syncing Stripe customer data into a CRM
- Preventing duplicate records across systems
- Handling ambiguous identity matches (needs_review escalation)
- Post-acquisition / multi-instance CRM cleanup (multiple Salesforce orgs, HubSpot portals, Stripe accounts)

## Key properties

- Read-only by design — ingests data, never writes to external systems
- Rules-based scoring — not an LLM guessing layer; ambiguous matches route to human review
- Three-tier matching pipeline: exact match, fuzzy reconciliation, human review
- Golden-record enrichment: entities accumulate identifiers over time from linked source records
- Multi-tenant with UUID primary keys and tenant-scoped data isolation
- Built-in OAuth connectors for Salesforce, HubSpot, Stripe (including multiple instances per provider)
- 11-tool MCP server for Cursor, Claude Desktop, and any MCP-compatible agent
- Matching analytics with tuning insights for continuous improvement

## Links

- Homepage: https://www.anchord.ai
- For Agents: https://www.anchord.ai/for-agents
- Scenarios: https://www.anchord.ai/scenarios
- Documentation: https://www.anchord.ai/docs
- Agent Quickstart: https://www.anchord.ai/docs/agent-quickstart
- Recipes: https://www.anchord.ai/docs/recipes
- API Reference (Swagger/OpenAPI): https://api.anchord.ai/docs
- OpenAPI spec: https://api.anchord.ai/api/openapi.yaml
- What is Anchord: https://www.anchord.ai/what-is-anchord
- FAQ: https://www.anchord.ai/faq
- Identity Resolution for AI Agents: https://www.anchord.ai/use-cases/identity-resolution-for-ai-agents
- Prevent Duplicate Records: https://www.anchord.ai/use-cases/prevent-duplicate-records
- Read-Only Trust Layer: https://www.anchord.ai/use-cases/read-only-trust-layer
- CRM Data Quality for Agents: https://www.anchord.ai/use-cases/crm-data-quality-for-agents
- Post-Acquisition Cleanup: https://www.anchord.ai/use-cases/post-acquisition-cleanup
- Sign up: https://app.anchord.ai/signup