Agents, Built for the Human + Agent Era

Your AI team, and
every surface it runs in.

iCustomer runs role-based AI agents, iWorkers, on your audience, and exposes governed agent surfaces, MCP Server, CLI, One API, and iConnect, so the agents you already use work with the same context and the same controls.

The iWorkers roster

Role-based teammates that execute decision loops. You set the goals; they run the cycles.

iWorker

Performance Marketing

Always-on media buyer inside your policy gates.

iWorker

Account Strategist

Runs account-based plays from live FIRE scores: fit, intent, recency, engagement.

iWorker

Demand Gen

Turns FIRE buying signals into pipeline plays.

iWorker

Content Strategist

Matches content to audience intent in real time.

iWorker

Commerce Growth

Drives conversions and repeat purchases for D2C.

iWorker

Retention Engine

Expands accounts and catches churn risk early.

iWorker

Data Steward

Keeps the loop's data clean and compliant.

iWorker

Compliance Guard

Enforces compliance across every activation.

Every surface agents run in

For teams and for code. One governed toolset behind all of them.

Surface

MCP Server

Audiences, FIRE scores, and decisions exposed as governed tools any agent can call.

Surface

iCustomer CLI

Ship loops as code from Claude Code or Codex. Config versioned in git.

Surface

One API

Every connected tool behind one endpoint, with PII and leak protection on every payload.

Surface

iConnect

Meta orchestrator: one router across every connected tool, inside the Harness guardrails.

Bring your own agents

If you can describe it in Claude Code or Codex, you can deploy it.

Agent client

Claude & Claude Code

Ask for audiences, FIRE scores, and launches over MCP, in chat or in code.

Agent client

OpenAI Codex

Codex agents deploy and operate loops through the same governed MCP tools.

Agent client

Google Gemini

Gemini agents call the same governed tools over MCP. Bring the model you run.

Model layer

iCustomer SLMs

Domain-adapted small models score and decide inside your cloud. Low inference cost, nothing raw leaves.

The context moat

Agents are becoming commodities. Context is not.

Every brand can run the same agents. Only yours run on your audience brain, grounded in your four context pillars, inside your guardrails, with every decision traced.

Governed by design

Enterprise trust is not a feature toggle. It is the architecture.

Governance

Harness

Guardrails, evals, and controls on every agent action. Agents propose and execute inside policy gates.

Governance

Decision Traces

Every action logged with who, why, and outcome. Human and agent activity, equally auditable.

Governance

Human-in-the-Loop

Your team sets goals, approves exceptions, and audits outcomes. Taste stays in the loop.

Governance

Consent on Every Payload

Activation never spends consent capital you have not earned. PII protected by design.

Agents, answered

iWorkers are iCustomer's role-based AI agents: teammates like a Performance Marketing buyer, an Account Strategist, and a Retention Engine that execute decision loops on your audience, inside Harness guardrails, with a Decision Trace on every action.

Yes. Claude, Claude Code, OpenAI Codex, and Google Gemini connect over the iCustomer MCP Server and call the same governed tools iWorkers use: audiences, FIRE scores, decisions, and activations. One surface, any agent.

Every agent action runs inside the Harness: guardrails, evals, and policy gates, with consent checked on every payload and a Decision Trace logging who acted, why, and what happened. Humans set goals, approve exceptions, and audit outcomes.

No. Agents execute; your team decides what matters. iCustomer is built human-in-the-loop: your team's taste and audience knowledge are the edge nobody can copy, and the loop is designed so that edge compounds instead of leaking.