Build with LLMs

Cleeng provides the subscription and identity model, allowing you to leverage vibe coding, or low-code AI app builders with our hosted experiences to quickly deploy sophisticated, AI-driven, and agentic solutions. The Cleeng MCP Server connects Cleeng to the AI tool your team already uses, so you can query subscriber data and request changes in conversation.

LLM-friendly documentation

We've made all our developer tutorials and API references available in a streamlined, machine-readable format for Large Language Models (LLMs). This accelerates your agent development:

• Direct LLM-friendly documentation - index of the guides and API reference pages.

Cleeng MCP Server

Model Context Protocol (MCP) is an open standard that lets an AI tool access external tools and data sources.

Cleeng builds and hosts the server, so there is nothing to deploy or maintain: you add the server URL in your AI tool, sign in with your Cleeng Dashboard login, and then ask for data or request changes in conversation. Cleeng provides the connection; you bring the AI account, and once data reaches that tool it is governed by your agreement with its provider.

Claude is supported today, and support for further tools is planned.

Before you connect an AI tool that supports MCP, you need a Cleeng Dashboard account (you cannot sign up from inside the AI tool) and MCP access granted to you by an administrator in the dashboard. Sign-in happens on Cleeng's own pages, so the AI tool never receives your password.

For connector settings and a walkthrough with screenshots, see Connect your AI tool.

What it can reach

Access is scoped by three roles, granted per user in the dashboard: Retention Intelligence, Customer Experience Management, and Product Management. An administrator grants each role at Read only or Read + write. At the consent screen you choose how much of that access to pass on to the AI tool, so a tool never holds more than you do, and revoking the authorisation stops further access. For the functions in each role, see What you can do with each role.

How it behaves

Lookups and metrics answer immediately and change nothing. Changes are previewed, and written only after you confirm in the conversation and Cleeng validates them against your account rules. If you set the AI tool permission level to Read, the tool describes a change but never writes it.

Figures come from what Cleeng already reports; Cleeng creates no new reporting for the AI tool.

Administrators: grant and manage Cleeng MCP access.

What the agents read from

An AI agent is only as accurate as the data behind it. Cleeng normalises subscriber data before an agent reads it. The model handles common subscriber-data problems, including:

• One subscriber, many channels. A viewer may register on web, purchase through an app store, and contact support through another system. Each source holds a fragment, and Cleeng is designed to resolve them into a single subscriber.
• Stores define an active subscription differently. Apple, Google, and Roku each draw the line between paid access and lapsed access in their own way. Cleeng maps these into one lifecycle.
• Events can arrive late, twice, or out of order. Store and payment notifications replay and correct themselves after the fact. Cleeng reconciles them into one record of what happened.

The MCP Server, the Cleeng dashboards, and data exports all read from this same model.

For how the data is ingested, modelled, and made available, see AI subscriber data.

AI-assisted integration examples

For rapid deployment by no-code, low-code, and full-stack teams, the fastest path to a complete subscriber journey (registration, checkout, and account management) is Cleeng Hosted Customer Flows.

See a few examples

• Claude.ai and Cleeng: Integrating Cleeng within your Claude App (video).
• Lovable and Cleeng: Integrating Lovable with Cleeng Hosted Widgets.
• Base44 and Cleeng: Integrating Base44 with Cleeng Hosted Widgets.

Architectural pattern

Let your AI or agentic layer focus on discovery, recommendations, and the user interface. Cleeng remains the authoritative System of Record for all core business functions related to subscriber management, including:

• Users / Subscribers
• Offers
• Billing and payments
• Subscriptions
• Access control and entitlements

Once this foundation is established, you can scale your integration with more advanced server-side connectors and custom agents to orchestrate Cleeng alongside your Content Management System (CMS), video player, or marketing and data platforms.

What stays your responsibility

Remember to always follow good practices when using AI.

• Grant least privilege. Give an AI tool the narrowest access that lets the task work, and check whether read access alone is sufficient.
• Keep a human in the loop for consequential changes. Every write through the MCP Server is previewed and applied only after you confirm it.
• Verify before you act. Check answers against Cleeng records and your existing reporting, and review AI output before acting on it.
• Keep credentials and sensitive data out of prompts. Never paste API keys or tokens into an AI tool, and share subscriber data only as far as the task needs.
• Review access regularly. Review connected AI accounts and authorisations, and revoke access promptly when someone changes role or leaves.

Full detail, including the security boundary and a pre-connection checklist: Trust and governance.

Learn more

• Cleeng MCP Server: what it is and how it works