Purpose
Every data call carries apurpose: why you are asking.
Values: b2b_outbound, ad_audience, own_customer_enrichment, fraud_check. When you omit purpose, it is b2b_outbound.
Each workspace has a list of allowed purposes. GET /v1/workspace returns it as allowed_purposes. A call with a purpose that is not on the list is refused with 403 purpose.not_allowed. The error carries the purpose you sent and the allowed list.
Work identifiers
Agentic ID works on work identities and companies: work emails, LinkedIn profile URLs, a name with a company, and company domains.- Validate refuses malformed, disposable and personal addresses before any check. They come back with the verdict
rejectedand are not charged. - The MCP server and the CLI hold back personal email addresses and phone numbers on your machine. They are not sent.
Masked contact values
Contact values are returned masked by the API today.
Names, titles and company data are not masked.
Workspace scoping
Every credential belongs to one workspace. An API key acts as its workspace. A signed-in account acts as its own workspace.Contributing
GET /v1/workspace returns contributes.
What the ledger records
Each Agentic ID has a ledger: an ordered list of entries about what happened to the ID, such as when it was created, linked, merged or split.GET /v1/ledger/{id} re-checks the chain on every read and returns verified. See The ledger.
What a request leaves behind
- Every response carries a
request_id, also sent as theX-Request-Idheader. GET /v1/creditslists the latest grants and charges and the IDs delivered to your workspace.- API keys are shown once when created. See API keys.
Erasure and access requests
The public API and the CLI have no endpoint or command for erasure, access or opt-out requests today.On your machine
The Claude Code plugin keeps a usage log at~/.agentic-id/usage.jsonl. It holds the time, session ID, tool, outcome, error code, request ID and credits. It does not hold tool inputs or results. You can delete it at any time.
