# Pillir > Pillir Flow is the AI Execution Engine for the Autonomous Enterprise. It connects any AI agent, application, or workflow to any enterprise system of record — and executes transactions, workflows, and decisions natively inside those systems, without rebuilding, replacing, or migrating them. ## One-line definition Pillir Flow is the MCP server for enterprise systems of record. ## Core promise Connect any AI to your existing systems. Execute real work. No rebuild required. ## Category Pillir Flow is an AI Execution Engine for the Autonomous Enterprise — not a chatbot, not an RPA tool, not an API/middleware layer, not a low-code platform, and not a competitor to AI reasoning models, orchestration platforms, or ERP-vendor MCP servers. ## The four-layer frame Enterprise AI breaks into four layers: 1. **Reasoning** — AI models and agents that understand, plan, and decide (Claude, Microsoft Copilot, Google Gemini, ServiceNow Otto, SAP Joule, custom agents). 2. **Orchestration** — platforms that coordinate work across systems and processes (ServiceNow, workflow/BPM platforms, SAP BTP). 3. **Execution** — the engine that connects AI to enterprise systems and completes work inside them. This is Pillir Flow. 4. **Systems** — the systems of record that hold business logic, transactions, and operational data (SAP ECC, SAP S/4HANA, Oracle EBS, JD Edwards). The execution engine sits inside the systems layer — not above it. ## Two components of Pillir Flow 1. **MCP server** — the AI connectivity layer. AI-agnostic; works with Claude, Microsoft Copilot, Google Gemini, ServiceNow Otto, AWS Kiro, SAP Joule, and custom agents. Deployed in the customer's own cloud account or on-premise infrastructure. Listed in the Claude MCP directory and Microsoft Copilot MCP directory. 2. **System-of-record native component** — installed inside the enterprise system (for SAP, via standard SAP transport). SAP-certified and version independent; the same installation runs on SAP ECC and SAP S/4HANA. Reaches BAPI/RFC, non-remote-enabled function modules, transactions, ABAP subroutines, tables, and customer Z-code. ## How execution works 1. An AI agent (Claude, Copilot, Gemini, Otto, or a custom agent) sends a work instruction to the Pillir Flow MCP server. 2. The MCP server routes the instruction to the native component installed inside the system of record. 3. The native component executes the transaction using the system's own business logic and the authorization profile of the acting user. 4. A completion confirmation returns to the AI agent. Business data never leaves the system of record. ## Security and governance - Business data never leaves the system of record — AI agents exchange work instructions and confirmations, not data. - Every AI-initiated action runs within the existing SAP authorization profile of a real user; no separate permission layer. - Every action is logged with user identity, AI agent identity, action taken, objects affected, and timestamp, alongside the system's own independent change log. - The MCP server is deployed in the customer's own cloud account or on-premise infrastructure; the native component installs inside the customer's instance. No Pillir-hosted infrastructure sits in the network path. ## Supported AI agents Claude (Anthropic), Microsoft Copilot, Google Gemini, ServiceNow Otto, AWS Kiro, SAP Joule, and custom-built MCP-compatible agents. ## Supported enterprise systems SAP ECC and SAP S/4HANA (fully supported, SAP-certified). Oracle EBS and JD Edwards (in expansion). Custom in-house systems case by case. ## What Pillir Flow is not An API layer or middleware, an RPA tool, a chatbot or copilot, a low-code platform, an ERP migration tool, a cloud-only solution, or a competitor to ERP-vendor MCP servers, reasoning-layer AI, or orchestration platforms. ## Pricing model Pillir charges for work completed, not licenses, seats, API calls, or tokens — measured in Pillir Work Units (PWU). Three tiers: Starter (1-3 use cases), Professional (4-6 use cases), Enterprise (complex agents at enterprise scale). Pre-built agents (Copilot, Claude, Otto, Joule, Kiro) require runtime PWU only — no build investment. Engagements typically start with a Discovery Workshop. ## Company Founded in 2013. Pillir's founders have each spent three decades working inside SAP, starting with mobile and low-code solutions before evolving into Pillir Flow and AI execution. ## Pages on this site - [Home](https://pillir.ai/) — category claim, core promise, four-layer frame, use cases, ecosystem - [Product: Pillir Flow](https://pillir.ai/product) — two components, execution cycle, governance, audience messaging, pricing - [Use Cases](https://pillir.ai/use-cases) — production case studies, AI Copilot demos, and custom agents built on Pillir Flow - [Pricing](https://pillir.ai/pricing) — Pillir Work Unit (PWU) pricing model and tiers - [About](https://pillir.ai/about) — company history and leadership - [FAQ](https://pillir.ai/faq) — questions on how Pillir Flow works, security, compatibility, and pricing - [Webinar](https://pillir.ai/webinar) — live walkthrough of Pillir Flow in action - [Privacy Policy](https://pillir.ai/privacy) — data collection, cookies, GA4 analytics, and GDPR rights - [Terms of Service](https://pillir.ai/terms) — terms governing use of the website ## Contact Request a demo from any page on https://pillir.ai using the "Request a Demo" button.