Skip to main content
Licentric

Use case · AI agent monetization

Per-seat pricing breaks when the user is a robot.

Licentric is the monetization platform for the AI era — agent identity, token budgets, usage metering, and MCP authorization in one API.

Autonomous agents do not behave like seats. They work 24/7, span multiple AI providers, and consume tokens that vary 375× in cost across models. The infrastructure to give every agent its own identity, its own budget, its own audit trail — and to bill customers in a model that scales with what their agents actually do — does not exist in your stack today.

The problem

The AI-era monetization gap is not theoretical.

Five numbers describe a stack that was not built for autonomous agents. Each statistic below is sourced from Licentric's internal market research in VISION.md §1 — we cite our own SSOT and label external attribution where it exists, rather than inventing analyst names.

50:1

Non-human identities outnumber humans in AI-enabled enterprises. Service accounts, agents, and machine credentials now dwarf the human user base — and each one needs an identity, a budget, and an audit trail your IdP was never built for.

Source: Licentric VISION.md §1

90%

Of AI agents hold 10× more privileges than they need. Least-privilege has never been retrofitted to autonomous agents. The blast radius of a compromised agent is the union of every scope it could have been granted.

Source: Licentric VISION.md §1

~1,000

MCP servers exposed to the internet with zero authentication. The Model Context Protocol shipped without a default auth layer; the ecosystem is racing to retrofit OAuth 2.1 on top of services that were already in production.

Source: Licentric VISION.md §1

Aug 2026

EU AI Act phased enforcement begins. The regulation mandates governance, risk classification, and human-oversight records for AI systems. Evidence retention is now an audit obligation, not a security best practice.

Source: Licentric VISION.md §1 & §12

<20%

Of enterprises are ready for AI governance (Deloitte 2026). The compliance, identity, and metering primitives required by the EU AI Act and analog US/UK guidance do not exist in most stacks today.

Source: Licentric VISION.md §1 (Deloitte 2026)

What Licentric provides

Pillar 2 — AI Agent Monetization

Each label reflects what's shipped today — verifiable against the live REST API, the Python and TypeScript SDKs, and the dashboard. The two remaining roadmap items are specific unshipped slices (multi-model per-provider cost attribution; the operator compliance dashboards) — not the underlying APIs, which have shipped. Pillar 5 (Truthfulness) governs every label below.

Phase 1 (built)

Agent identity

Per-agent credentials with full lifecycle — issue, suspend, reinstate, revoke, and rotate. Each agent has a stable identity separate from the human user who created it. API: /api/v1/agents and /api/v1/agents/[id]/credentials.

Phase 1 (built)

Token budgets

Hard ceilings per agent, per team, or per workflow — enforced server-side via Redis counters, not client-side hints. Webhook fires when an agent hits its budget. API: /api/v1/agents/[id]/budgets.

Phase 1 (built)

Usage metering

Ingest token-events, call-events, and outcome-events at scale. Buffered with idempotency keys; flushes to durable storage on a deterministic cadence. API: /api/v1/metering/events and /api/v1/metering/usage.

Phase 1 (built)

MCP auth layer

Scoped, short-lived credentials for MCP servers via a token-exchange API — issue, list, and revoke agent↔MCP-server connections to fill the ~0%-default-auth gap in the protocol. Shipped: API /api/v1/mcp/* + SDK client.mcp. (Full OAuth 2.1-conformant flow + scoped audience tokens continue to mature.)

Phase 1 (built)

Outcome metering

Track what agents accomplished — tasks completed, meetings booked, issues resolved — alongside what they consumed, with idempotency-key deduplication and a completion lifecycle. Shipped: API /api/v1/outcomes/* + SDK client.outcomes.

Phase 3 (roadmap)

Multi-model cost attribution

Cost attribution per meter and per agent is shipped — API /api/v1/analytics/costs + SDK client.analytics + the Agent Costs dashboard. The remaining roadmap slice is multi-MODEL per-provider attribution (spend split across OpenAI / Anthropic / Google), which needs the per-model price spine populated.

Phase 3 (roadmap)

EU AI Act compliance dashboards

The compliance APIs are shipped — an append-only audit trail, risk assessments, and an Article-12 evidence-export bundle (/api/v1/compliance/*). The remaining roadmap slice is the operator-facing compliance DASHBOARDS, ahead of the August 2026 EU AI Act enforcement deadline.

Hybrid pricing models

Four pricing shapes that survive contact with autonomous agents.

Tokens vary 375× in cost across models. Agent utilization varies by orders of magnitude across customers. A monetization model that fits human seats wildly mis-prices both ends of the distribution. Below are the four shapes that compose every working AI-product pricing page we have studied.

Pure subscription

Flat monthly fee per agent or per workflow. Predictable revenue; predictable customer cost. Works when agent utilization is roughly uniform across customers.

$49/mo per active agent. Subscription drives access; token consumption inside the budget is included.

Subscription + usage overage

Flat fee includes a token-budget allowance; usage beyond the allowance is metered per million tokens. Most flexible — pairs predictable revenue with high-growth customer upside.

$49/mo includes 5M tokens per agent. Overage: $0.50 per additional 1M tokens, billed monthly.

Outcome-based

Charge per task completed, per meeting booked, per issue resolved. The agent's value is denominated in the customer's business unit, not the platform's tokens.

$1.20 per qualified meeting booked. Tokens consumed inside the workflow are absorbed by the platform.

Hybrid (subscription + outcome + overage)

Subscription covers access. Outcome events are billed when they hit a defined trigger. Token overage protects margin if a customer drifts toward unprofitable agents.

$199/mo base + $0.80 per qualified outcome + $0.40 per additional 1M tokens above 10M / agent.

Code

License keys and the agent control plane — one SDK.

The Python and TypeScript SDKs (v0.3.0) ship the full surface: license-key validation for your human users, plus agent identity, token budgets, metering, registry, fleet, evals, observability, governance, and cost. The REST API is the source of truth; the SDK is a thin, typed wrapper over it.

Python — shipped today

# Shipped today in licentric==0.3.0 (PyPI) — license keys AND the agent control plane.
from licentric import Licentric

client = Licentric(api_key="lk_live_...")

# License validation for human users.
result = client.validate(key="ACME-XXXX-XXXX-XXXX-XXXX", product_id="prod_123")

# Give an autonomous agent its own identity.
agent = client.agents.create(name="support-triage-agent", type="api_consumer")

# Cap its monthly token budget — enforced server-side.
client.budgets.create(
    agent.id,
    budget_type="tokens",
    limit_value=5_000_000,
    period="monthly",
)

# Meter what it consumes.
client.metering.ingest([
    {
        "agentId": agent.id,
        "eventType": "chat.completion",
        "meterName": "tokens",
        "quantity": 12_400,
        "metadata": {"model": "claude-sonnet-4-5", "workflow": "ticket-triage"},
    }
])

TypeScript — shipped today

// Shipped today in @licentric/sdk 0.3.0 (npm) — license keys AND the agent control plane.
import { Licentric } from "@licentric/sdk";

const client = new Licentric({ apiKey: process.env.LICENTRIC_API_KEY! });

// Give an autonomous agent its own identity.
const agent = await client.agents.create({
  name: "support-triage-agent",
  type: "api_consumer",
});

// Cap its monthly token budget — enforced server-side.
await client.budgets.create(agent.id, {
  budgetType: "tokens",
  limitValue: 5_000_000,
  period: "monthly",
});

// Meter what it consumes.
await client.metering.ingest([
  {
    agentId: agent.id,
    eventType: "chat.completion",
    meterName: "tokens",
    quantity: 12_400,
    metadata: { model: "claude-sonnet-4-5", workflow: "ticket-triage" },
  },
]);

// ...and the rest of the control plane: client.fleet, client.evals,
// client.observability, client.agentPolicies, client.analytics.

Why this matters now

Regulation and market timing are both pulling forward.

Market

Per Licentric VISION.md §4: the AI agent infrastructure market is projected to grow from $7.8B (2025) toward $52B (2030) at a 46% CAGR. The AI governance + compliance market is projected from $2.5B (2025) toward $68.2B (2035) at a 39.4% CAGR. Licentric's combined addressable market expands roughly 30-50× by 2030 relative to the licensing-only prior framing.

Regulation

EU AI Act phased enforcement begins August 2026. Risk classification, human-oversight records, and audit-trail retention become statutory — not best-practice. The compliance scaffolding (audit log immutability, evidence tables, per-agent risk classification) is built into Licentric today; the regulator-export views and operator dashboards land in Phase 3 ahead of the deadline.

AI Builder Beta Program

Free access for 10-20 companies building AI agents.

The AI Builder Beta Program offers free access in exchange for feedback and case studies. First 50 customers receive founder-direct migration assistance and a 30-minute fit-check call. Email [email protected] with a one-paragraph description of what you're building, the agent count you expect to hit, and which AI providers you depend on.

Frequently asked questions

Honest answers about what ships today.

How is this different from Stripe metering or Paid.ai?

Stripe meters billing events but has no concept of agent identity, token budgets, or MCP authorization. Paid.ai focuses on outcome-billing + value receipts but ships no license keys, no agent-identity primitives, no MCP auth, and no per-agent token-budget enforcement. Licentric covers identity (who is the agent?), authorization (which MCP servers and what scopes?), budgets (what is its spending ceiling, server-enforced?), AND metering — alongside the license-key stack you already need for human users.

Why is per-seat licensing broken for AI agents?

Autonomous agents are not seats. One human can deploy one agent or one hundred. Agents work 24/7, not the 40-hour week per-seat pricing assumed. Token consumption varies 375× across models. A per-seat SaaS plan that fits human users wildly over-charges low-volume agent customers and wildly under-charges high-volume ones. Hybrid pricing (subscription + usage + outcome) is the only model that scales with how agents actually behave.

What's actually shipped today vs roadmap?

Shipped today, each with API + SDK + dashboard: agent identity (CRUD + suspend/reinstate/revoke/rotate credentials), server-enforced token budgets, usage metering, scoped MCP credentials, and outcome metering — plus the full control plane: registry (versioning + one-click rollback), fleet management (batch lifecycle/rollout/rollback), evaluations (record/score/compare), observability (redacted trace ingest + span-waterfall replay), and cost attribution. Governance (an allow/deny policy engine at the agent tool-call seam) is shipped but opt-in/default-off. Coming next: human-in-the-loop approvals for high-consequence actions. On compliance we make no 'EU AI Act compliant' claim — governance supports the Article 12/14 evidence obligations where agents are high-risk-classified, and an Article-12 evidence-export endpoint is available.

Are SDKs available for the agent / control-plane APIs?

Yes. The Python SDK (licentric on PyPI, v0.3.0) and the TypeScript SDK (@licentric/sdk on npm, v0.3.0) both ship the full surface — license-key helpers (validate, activate, deactivate, heartbeat, getLicense, checkout, offline verification) AND the agent control plane: client.agents (identity, credentials, versions, rollback), client.budgets, client.metering, client.fleet, client.evals, client.observability, client.agentPolicies, and client.analytics. The REST API is the source of truth; the SDK is a thin, typed wrapper over it.

What compliance does Licentric have for AI agent monetization?

GDPR Article 17 (erasure) and Article 20 (data portability) are verified per release — agent identities, credentials, budgets, and metering events all flow through the deletion and export RPCs. SOC 2 Type II is pursuing Q4 2026 (not yet certified). ISO 27001 is not pursued today. EU AI Act readiness scaffolding (audit logs, risk classification tables, compliance evidence migration files 030+) is built today; the operator dashboards and the regulator-export views are Phase 3. We do not advertise any compliance posture we do not hold — Pillar 5 (Truthfulness).

How do I join the AI Builder Beta Program?

Email [email protected] with a one-paragraph description of what you're building, the agent count you expect to hit, and which AI providers you depend on. 10-20 companies building AI agents get free access in exchange for feedback and case studies. The program is founder-direct; first 50 customers receive migration assistance and a 30-minute fit-check call.