The History of MCP and ACP: Where Did These Ideas Come From and Who’s Driving Adoption?
In the past year, two acronyms have quietly rewritten the playbook for enterprise AI: Model Context Protocol (MCP) and Agentic Commerce Protocol (ACP). While one connects AI agents to enterprise systems and data, the other enables AI-driven shopping and transactions.
Together, they hint at something bigger — a world where AI isn’t just conversational, it’s actionable. But where did these protocols come from, and why are they so central to the next era of enterprise technology?
Let’s trace the origins of both, understand the forces behind them, and explore why this shift matters to search, discovery, and decision-making platforms like Lucidworks.
1. Setting the Stage: Why Protocols Became Necessary
Before MCP or ACP, AI systems lived in silos. Each large language model (LLM) — OpenAI’s GPT, Anthropic’s Claude, Google’s Gemini — had to be custom-integrated with every API, tool, or data source it wanted to access.
That meant building one-off connectors, brittle integrations, and high-maintenance middleware. For enterprises with complex data architectures — CRMs, ERPs, search indexes, knowledge bases — it was unsustainable.
The AI community quickly recognized a need for standardization, similar to what HTTP did for the web or USB-C did for hardware.
That’s where Model Context Protocol (MCP) entered the picture.
2. The Birth of MCP: From Closed Models to Connected Ecosystems
MCP, short for Model Context Protocol, was introduced by Anthropic and open-sourced in late 2024.
Its goal was deceptively simple: give AI models a structured, secure way to “plug in” to external tools and data systems.
In other words, MCP is to AI agents what APIs are to applications — a universal language for exchanging context and capabilities.
2.1 What MCP Does
MCP defines:
- How AI clients (like an LLM) request and receive contextual data
- How servers (like enterprise systems) expose capabilities, actions, and data
- How both parties enforce security, governance, and access control
This architecture mirrors how a web browser and web server communicate — except instead of HTML and HTTP, it’s JSON-RPC and typed schemas governing what the AI can access, invoke, and return.
2.2 Early Adoption and Governance
Shortly after its release, Red Hat, AWS, Solo.io, and IBM began experimenting with MCP extensions.
- AWS highlighted MCP’s potential to standardize AI tool invocation across the cloud.
- IBM framed it as an enabler for “AI governance and transparency.”
- Microsoft integrated MCP concepts into Copilot Studio, its enterprise orchestration environment.
Academic researchers also began auditing MCP for security, maintainability, and policy compliance, signaling it had moved from concept to infrastructure.
The open-source repository on GitHub, maintained by Anthropic and contributors, became the home for MCP SDKs, schemas, and community extensions.
In short, MCP became the connective tissue of agentic AI, defining how context moves between models and systems.
3. The Rise of ACP: Commerce Gets Its Own Protocol
While MCP focused on data and tool access, another challenge was brewing: how could AI safely transact?
When OpenAI announced that ChatGPT could now “buy” products using built-in payment capabilities, it wasn’t just a product update — it was the debut of a new standard called the Agentic Commerce Protocol (ACP).
3.1 What ACP Is
ACP stands for Agentic Commerce Protocol, a standard that governs how AI agents initiate and complete e-commerce transactions.
Where MCP deals with “read and reason,” ACP deals with “buy and act.”
ACP defines:
- How agents securely request product data, pricing, or availability
- How checkout, payment, and fulfillment are handled under strict merchant control
- How consent, encryption, and payment tokens are managed to ensure security and compliance
In simple terms: ACP is the “checkout button” for agentic AI — a safe, standardized way to complete purchases initiated through AI assistants.
3.2 The Origin and Backers of ACP
ACP was jointly introduced by OpenAI and Stripe in early 2025.
- OpenAI wanted a way for ChatGPT to engage in transactions without holding payment data or becoming the “merchant of record.”
- Stripe saw an opportunity to extend its infrastructure to autonomous agents.
The model is simple yet transformative:
- The user provides consent and payment details securely.
- The agent transacts via ACP, under a scoped and auditable protocol.
- The merchant fulfills the order as usual — preserving existing business systems.
Over time, other commerce platforms began evaluating ACP as the baseline protocol for AI-assisted purchasing, especially for B2B marketplaces and enterprise procurement systems.
4. Timeline: From Proprietary Integrations to Open Protocols
| Year | Milestone | Key Organizations Involved |
|---|---|---|
| 2023 | Rise of multimodal and agentic AI capabilities | OpenAI, Anthropic, Google, Meta |
| 2024 | Early calls for AI interoperability and tool standardization | AI infrastructure startups, OSS community |
| Late 2024 | MCP (Model Context Protocol) was introduced | Anthropic, Red Hat, AWS |
| Early 2025 | ACP (Agentic Commerce Protocol) launched | OpenAI, Stripe |
| 2025 | MCP adoption in enterprise AI and developer ecosystems | IBM, AWS, Microsoft, Solo.io |
| 2025+ | Integration of MCP + ACP patterns in search and commerce ecosystems | Lucidworks, commerce platforms, enterprise data vendors |
5. Why MCP and ACP Matter to Enterprises
Both MCP and ACP address the same foundational challenge: how to connect reasoning systems to the real world, safely and scalably.
For enterprises, this is where AI shifts from “assistant” to “operator.”
- MCP gives AI agents controlled access to internal systems like CRMs, product catalogs, or document stores.
- ACP gives AI agents the means to execute commercial actions — place orders, negotiate pricing, or reorder inventory.
Together, they create a secure ecosystem where context becomes the fuel and transactions become the outcome.
6. Connecting to Lucidworks: Search and Discovery in the MCP Era
Lucidworks has long built solutions that turn data into decisions — combining powerful search, recommendation, and signal intelligence with enterprise-grade governance.
In the MCP era, Lucidworks’ role expands from indexing and ranking to context provisioning:
- MCP allows AI agents to dynamically query Lucidworks indexes.
- Lucidworks’ Signals framework can feed those agents a rich behavioral context.
- The Lucidworks Platform acts as an MCP server, exposing search and discovery capabilities as callable resources.
In a commerce context, MCP and ACP can work hand-in-hand:
- MCP helps the AI understand what’s in the catalog and what’s relevant to the user’s intent.
- ACP enables the final step — completing the transaction within approved enterprise workflows.
Lucidworks effectively becomes the connective context layer that makes agentic AI safe, relevant, and auditable in enterprise environments.
7. How MCP and ACP Complement Each Other
| Aspect | MCP (Model Context Protocol) | ACP (Agentic Commerce Protocol) |
|---|---|---|
| Purpose | Connects AI to data, tools, and APIs | Enables AI to conduct secure transactions |
| Creator | Anthropic (open-source, 2024) | OpenAI + Stripe (commerce-focused, 2025) |
| Scope | Knowledge retrieval, workflow orchestration | Payments, procurement, fulfillment |
| Architecture | JSON-RPC over secure channels | API and token-based commerce calls |
| Enterprise Relevance | Search, knowledge management, automation | E-commerce, B2B transactions, supply chain |
| Lucidworks Role | Provides contextual and indexed data for MCP agents | Enhances discovery before ACP checkout actions |
8. Practical Example: An Enterprise Buyer with MCP + ACP
Let’s imagine a B2B purchasing agent for a manufacturing firm.
- Query (via MCP): The buyer’s AI agent asks: “Find suppliers of heat-resistant composite materials with delivery under 5 days.” Lucidworks, acting as an MCP server, surfaces supplier data, past performance, and pricing signals.
- Contextual Refinement: The agent filters results based on compliance and inventory data pulled through other MCP servers.
- Negotiation & Checkout (via ACP): Once a supplier is chosen, ACP takes over — securely handling negotiation, PO submission, and payment authorization under enterprise control.
- Audit & Governance: Both MCP and ACP generate logs for transparency — every decision and transaction is recorded for compliance and policy review.
This combined flow illustrates why protocols, not bigger models, are driving the next phase of AI evolution.
9. Who’s Driving Adoption Next?
Beyond the originators, an emerging ecosystem is accelerating MCP and ACP adoption:
- Cloud providers (AWS, Azure, GCP) are building managed protocol gateways.
- Open-source communities are expanding MCP server templates and SDKs.
- Enterprise AI platforms like Lucidworks are incorporating MCP-native integrations for contextual retrieval and governance.
- Fintech and commerce vendors are exploring ACP extensions for fraud detection and returns automation.
By mid-2026, many analysts expect MCP and ACP to become default AI infrastructure standards — the same way REST and OAuth did for web services.
10. Looking Ahead: Protocols as the AI Operating System
If LLMs are the brains of AI systems, protocols like MCP and ACP are their nervous system — defining how they sense, act, and communicate.
In the enterprise, that means:
- Better interoperability across teams and tools.
- Transparent, governed access to sensitive systems.
- Richer personalization powered by real-time context.
- Safe, compliant AI transactions — from search to checkout.
Lucidworks sits at the center of this transition, helping it happen to usher in an even bigger future. Its expertise in signal intelligence, contextual search, and enterprise architecture makes it a natural bridge between LLMs and the systems that matter most.
Key Takeaways
- MCP (Model Context Protocol) was created by Anthropic in 2024 to standardize how AI agents connect to external tools and data.
- ACP (Agentic Commerce Protocol) was introduced by OpenAI and Stripe in 2025 to enable safe, governed AI-driven transactions.
- These protocols shift focus from bigger models to smarter, safer, more interoperable AI systems.
- Lucidworks provides the contextual intelligence that powers both, enabling enterprises to adopt MCP and ACP confidently.
- The future of enterprise AI will be protocol-driven — where search, context, and action converge through standards that make AI truly operational.