Built for GTM Agents & Builders

One open ID for every person, company and agent.

For the AI agents that run your go-to-market, and the people who build them.

Resolve any person or company to one stable ID, with the evidence behind the match. Use it from Claude Code, Codex, your own agent, a CSV or a data pipeline.

Get free creditsView on GitHubView on GitHub
  • 75M+ people and companies resolved and verified
  • Used by 500+ teamsOpen spec, open code, community ledger
Install in Claude Code · 30 seconds
$ claude plugin marketplace add icustomerai/agentic-id
$ claude plugin install agentic-id@agentic-id
Works with
Five tools

Resolve is the core. Everything else builds on it.

Enrich, find and discover work on resolved identities. Validate checks a mailbox before you send.

Agentic ID
did:agenticid:person:7a1e…
active

Identify the people and companies on your site with iReveal. Send audiences to ad platforms and watch for changes with Growth Brain. Same IDs, same credits.

Built for how you work

One API for builders, agencies and data teams.

GTM buildersPoint the CLI at a CSV, set a budget, get rows back.
AgenciesRun each client in its own workspace, with its own credits and caps, so spend and results never mix.
Data and engineering teamsAPI keys for CI and pipelines, idempotent retries, clear exit codes and a dependency-free TypeScript SDK.
Works where your agent works

Same API, same answers, every surface.

Every surface takes the same inputs and returns the same outputs, error codes and credits. Use the MCP server where your agent supports it, and the CLI where it doesn’t.

Hard budgets on every call

Every call carries a credit cap. A call that can’t fit spends nothing and says so.

You pay for usage

No charge for misses. Rejected input and errors on our side are free too. Every answer shows the most it could have cost and what it did cost.

Errors an agent can fix

Every error has a machine-readable code and a plain fix the agent can act on. Retries with the same key never charge twice.

Claude Code
$ claude plugin marketplace add icustomerai/agentic-id
$ claude plugin install agentic-id@agentic-id
✓ Maya Chen · VP Marketing · Tallpine Software
id did:agenticid:person:7a1e0c9f3bqk2m8xn4vd6hwt0r
matched on work_email · confidence 0.97 · free
Marketing leaders at tallpinesoftware.com (2 of 7 found):
Daniel Okoro  Head of Demand Gen  unverified
Sofia Marin  Dir. Product Marketing  unverified
Charged 4 credits (cap 6).
Evidence, not guesses

Most data tools look things up. Agentic ID resolves.

Same person, any input, one ID. Every link needs evidence.

  1. How a match is decided

    Any input, one Agentic ID.

    A work email, a LinkedIn URL or a name + company gives one ID, matched on work_email, confidence 0.97.

  2. Strong evidence

    One stable ID, every time.

    Every issued ID is recorded.

  3. Weak input

    needs_evidence, with what to add.

    Bad input is refused before it costs anything.

  4. Two records

    Conflict, flagged, never merged.

    Every link needs evidence.

Lookup and enrichment toolsAgentic ID
Maya ChenMerged silentlyConflictFlagged, never merged
What you getA best guess for each requestOne stable ID, with what it matched on and how confident it is
Same person, two inputsTwo recordsThe same ID, whichever identifier you start from
Spend controlA monthly plan or a balanceA hard cap on every call and every file
What you pay forDepends on the vendorUsage only. No charge for misses
When you leaveThe vendor keeps the IDsYou keep your ID mappings, and the spec is open
The ledger

A shared ledger. Hashed links, never raw data.

Agentic ID is an open ledger protocol. iCustomer runs the first node. The reference node is open source, so anyone can run one.

Look up

Resolve any identifier against the shared ledger.

Contribute

Add hashed keys and the evidence behind a link.

Keep it private

A workspace can opt out of contributing and keep its keys private.

Who controls what

  • People and companiesOpt in or out with the brand.
    Make their choices
  • Your brandOwns first-party data, consent and ID mappings. Runs the opt-out.
    Controller
  • iCustomerLinks them with hashed IDs. Applies your consent first. Sells no data.
    Processor · neutral
  • Data sourcesSecond- and third-party attributes, supplied under license.
    Licensors

Ledger

Hashed keyEvidenceStatus
h(02bd..)consentopt-out applied
h(7c1e..)work_emaillinked
h(a94f..)LinkedIn URLlinked
h(b3d0..)name + companylinked

No brand relationship? People opt out with the data source that supplied their data.

No one owns a person.Agentic ID is a reference,not a record.

Why it’s open
  • Identity shouldn’t be rented.
  • The spec, the code and the ledger are public.
  • Any brand can check how an ID is made, and keep it.
How this started

We built it for ourselves first.

“So we built an ID system for ourselves. Identity shouldn’t belong to one vendor, so we’re opening the spec, the code and the ledger.”

We were building Growth Brain for growth teams, and every decision broke at the same step: one person showed up as five records across a CRM, a data cloud and three ad platforms.

It has since resolved and verified 75M+ people and companies for 500+ teams.

Where your data goes

Work identity today. Consumer identity next, with consent.

Work identifiers only, for nowWork emails, LinkedIn profile URLs, names with a company, and company domains. Personal emails, phone numbers and device IDs are refused until consumer identity ships with consent controls.
People in the US, companies everywherePerson lookups outside the US are refused. Company lookups work in every region.
Matching on hashed keysThe ledger stores hashed links, never raw data. Requests go only to the server you configured, and results come back into your agent’s conversation.
Open source

Agentic ID

  • Spec, SDK, CLI, plugin and reference node
  • Run it yourself
FreeView on GitHub
Hosted

Agentic ID Cloud

  • The shared ledger
  • Resolution across 100+ independent sources
Pay per useGet free credits
Enterprise

iCustomer Enterprise

  • Agentic ID inside your own data cloud
  • Ongoing monitoring, SLA and DPA, plus Growth Brain
AnnualBook a demo

Agentic ID questions

Agentic ID is an open identity protocol and a shared ledger. It resolves any person or company to one stable ID, with the evidence behind the match. It is built for GTM agents and builders, who use it through a Claude Code plugin, an MCP server, a CLI, an SDK or a REST API.
Every answer says what it matched on and how confident it is. Weak input comes back as needs_evidence, with what to add. Identifiers that point at two records come back as a conflict, flagged and never merged.
Four parties, four roles. People and companies make their choices with the brand. Your brand owns its first-party data, consent data and ID mappings, runs the opt-out process, and keeps everything if it leaves. Second- and third-party sources supply attributes under license. iCustomer is a neutral identity between the two, acting as your processor.
Through the brand. An opt-out is captured in the brand’s own consent tool and first-party data. Agentic ID reads that choice before any match or activation, so the person is left out everywhere the ID is used. iCustomer runs no separate opt-out and keeps no separate consent. Someone who has no relationship with a brand opts out with the data source that supplied their data.
The same person or company resolves to the same Agentic ID wherever they appear. The ledger holds hashed links only. No brand can see another brand’s records, audiences or results, and a workspace can keep its keys private.
Yes. The spec is licensed CC BY 4.0. The Claude Code plugin, CLI and SDK are MIT. The reference ledger node is AGPL-3.0. iCustomer runs the first node and the hosted edition. The code is at github.com/icustomerai/agentic-id.
Work identity for people in the US, and companies worldwide. Consumer identifiers, such as personal email, phone and device, come next and will be consent-gated. Until then they are refused.
Through the MCP server, Agent Skills, the CLI, the SDK or the REST API. Every call has a hard credit cap, and every error comes with a fix the agent can act on.
Any agent that speaks MCP can use the Agentic ID MCP server, and any language can call the REST API. Agent Skills work in any agent that reads SKILL.md files. Setup notes for Codex, Cursor and other agents are in the docs.
Yes. Give each client its own workspace, with its own credits and budget caps, so spend and results never mix. Every call is logged with its cost.
Yes. Use an API key instead of browser sign-in. The CLI reads CSV files, caps the budget for the whole file, retries safely with an idempotency key and returns clear exit codes. Rows it can’t send come back marked, never dropped.
Yes. The SDK, CLI and Claude Code plugin are MIT licensed. The reference ledger node is AGPL-3.0, and a different license is available on request.
No. Agentic ID is one of three ways into Growth Brain, the iCustomer platform, and they share one credit system. It is the way in for builders, and the same IDs power iReveal and the full platform.

Resolve your first record.

Install the plugin or call the API in a few minutes.New accounts get free credits.

Get free creditsView on GitHub
  • Priced on usage, not seats
  • One credit system
  • SOC 2 Type II