MCP Adoption Timeline: From Anthropic Experiment to Linux Foundation Standard

The Model Context Protocol hit 97 million SDK downloads per month before most enterprise teams had heard of it. That's faster than React's adoption curve - and React had a decade's head start.

This is the full MCP adoption timeline: every inflection point, every stat that matters, and what it means if you're building AI agents into SaaS products right now.

---

TL;DR

• July 2024: Two Anthropic engineers start building MCP.

• November 25, 2024: MCP launches publicly as an open standard.

• Early 2025: OpenAI, Google DeepMind, and Microsoft adopt it.

• December 9, 2025: Anthropic donates MCP to the Linux Foundation's Agentic AI Foundation (AAIF).

• Mid-2026: 21,000–23,000+ public MCP servers. 41% of enterprise software orgs in production.

• Bottom line: MCP is no longer a bet. It's infrastructure.

---

What Is MCP? The 60-Second Explainer

MCP - Model Context Protocol - is an open standard that lets AI agents connect to external tools, data sources, and APIs through a single, consistent interface.

Think of it as USB-C for AI agents. Before MCP, every tool integration was custom-built. One connector for Slack, another for GitHub, another for your internal database. Each one fragile. Each one maintained separately.

MCP replaces that with a universal protocol. An AI agent that speaks MCP can connect to any MCP server - instantly, without custom glue code. (For a full primer, see what is Model Context Protocol.)

Three things MCP defines:

• Resources: data the AI can read (files, database rows, API responses)

• Tools: actions the AI can take (run a query, send a message, create a ticket)

• Prompts: reusable interaction templates that standardize how agents behave

The spec is open. The SDKs are free. And as of December 2025, it's governed by the Linux Foundation - not any single vendor.

---

The MCP Adoption Timeline: Month by Month

The MCP adoption timeline is one of the fastest standardization arcs in developer infrastructure history. Here's the full picture.

Date	Event
July 2024	Anthropic engineers David Soria Parra and Justin Spahr-Summers begin development
November 25, 2024	MCP publicly launched as an open standard
Early 2025	OpenAI, Google DeepMind, Microsoft adopt MCP
May–Oct 2025	Remote MCP server adoption grows nearly 4×
Sep–Nov 2025	MCP server registry grows 407%
November 25, 2025	First MCP anniversary + spec version 2025-11-25 released
December 9, 2025	Anthropic donates MCP to the Linux Foundation's AAIF
Mid-2026	21,000–23,000+ MCP servers across all directories

The First 90 Days (Nov 2024 – Feb 2025)

MCP launched quietly. No splashy conference keynote. Just a GitHub repo, a spec, and two SDK libraries.

Within weeks, the developer community ran with it. Early adopters built connectors for GitHub, Postgres, and filesystem access. The PulseMCP registry didn't exist yet - but the momentum was already building.

The Breakout Quarter (Q2–Q3 2025)

This is when the MCP adoption timeline bent sharply upward.

• Remote MCP server adoption grew nearly 4× between May and October 2025.

• The server registry expanded 407% between September and November 2025.

• PulseMCP hit 5,500+ registered servers by October 2025.

The top three most-searched MCP servers by late 2025: Playwright MCP (35,000 monthly searches), Figma MCP (23,000), and GitHub MCP (17,000). All three are B2B tools. That's not a coincidence.

The Governance Moment (Nov–Dec 2025)

On November 25, 2025 - exactly one year after launch - Anthropic released spec version 2025-11-25. It added async task workflows, OAuth machine-to-machine flows, parallel tool calls, and a standardized .mcpb bundle format.

Two weeks later, on December 9, 2025, Anthropic donated MCP to the Linux Foundation. The protocol became a founding project of the Agentic AI Foundation (AAIF), alongside Block's goose and OpenAI's AGENTS.md.

Mid-2026: Scale

By mid-2026, the numbers are hard to argue with:

• 21,000–23,000+ public MCP servers across all directories

• 97M–110M SDK downloads per month (Python + TypeScript combined)

• ~67M local MCP server downloads in April 2026 alone

• ~18M server directory views per week

---

Why Did Anthropic Build MCP? The Problem It Solves

The problem was integration hell. Every AI agent needed custom connectors to every tool it touched.

Before MCP, building an AI agent that could read a Jira ticket, query a database, and post a Slack message meant writing three separate integrations. Each one broke when the upstream API changed. Each one was a maintenance burden.

The Model Context Protocol history starts with that frustration. David Soria Parra and Justin Spahr-Summers began development in July 2024 specifically to solve the N×M integration problem - where N agents and M tools produce an explosion of custom connectors.

MCP collapses that to N + M. Build once to the protocol. Connect to everything.

Three things that changed overnight once MCP launched:

01. Tool integrations went from weeks of custom work to hours of configuration.

02. Agent behavior became portable - the same agent could run against different MCP servers without modification.

03. Security and auth became standardized, not ad-hoc.

---

The Tipping Point: When Every Major AI Lab Said Yes

The MCP standard for AI agents became inevitable the moment OpenAI and Google DeepMind adopted it in early 2025.

That's the real inflection point in the MCP adoption timeline. Anthropic built it. But when its two biggest competitors endorsed the same spec, MCP stopped being "Anthropic's protocol" and started being the protocol.

Microsoft followed. So did Cloudflare, AWS, and Bloomberg - all as Platinum-level supporters of the AAIF by December 2025.

Why did they all say yes?

• No vendor lock-in. The spec is open. No one owns the protocol.

• Network effects. Every new MCP server is available to every MCP-compatible agent, regardless of which AI lab powers it.

• Developer momentum. SDK downloads reached 97M/month faster than React reached equivalent adoption. Labs don't fight developer momentum - they follow it.

By mid-2025, 93% of MCP servers were using Streamable HTTP transport, signaling a maturing, standardized deployment pattern. This wasn't experimentation anymore. It was production infrastructure.

---

From Experiment to Standard: The Linux Foundation Move

On December 9, 2025, Anthropic donated MCP to the Linux Foundation - and that changed everything about how enterprises evaluate it.

The Agentic AI Foundation (AAIF) is a directed fund under the Linux Foundation. MCP joined as a founding project, governed under Apache 2.0 (code) and CC BY-4.0 (documentation). The trademark "MCP" and "Model Context Protocol" now fall under Linux Foundation trademark policy.

Co-founders of the AAIF: Anthropic, Block, and OpenAI. Platinum supporters: AWS, Google, Microsoft, Cloudflare, Bloomberg. (We dig into what Linux Foundation MCP governance means for the protocol's future.)

Why does this matter for enterprise buyers?

Three reasons:

01. Vendor neutrality is now structural, not just promised. No single company controls the roadmap.

02. Procurement gets easier. Enterprise legal and security teams know how to evaluate Linux Foundation projects. The governance model is familiar.

03. Long-term stability is guaranteed. The Linux Foundation doesn't abandon projects. MCP has the same institutional backing as Kubernetes and Node.js.

The Anthropic MCP Linux Foundation move wasn't a handoff. Anthropic continues to invest in MCP development and participates actively in the project. It was a signal: this protocol is too important to belong to any one company.

---

MCP Enterprise Adoption: Where Things Stand in 2026

41% of enterprise software organizations are already in limited or broad production with MCP. That's from Stacklok's State of MCP in Software 2026 survey - not a vendor projection, a real measurement.

Here's the full breakdown:

Stage	% of Software Orgs
Planning or evaluating	29%
Pilot	30%
Limited production	29%
Broad production	12%

Who's Building MCP Servers?

• 70% of MCP servers come from B2B companies.

• 80% of the top 20 most-searched MCP servers offer remote deployment - the enterprise-preferred model.

• 93% of servers use Streamable HTTP transport.

What's the Business Impact?

One case study shows a 700% increase in tool usage after integrating agents with MCP. That's not a rounding error. That's a step-change in what agents can actually do inside a product.

And 72% of MCP builders expect their usage to increase over the next 12 months, according to Zuplo's State of MCP report.

The MCP enterprise adoption curve isn't flattening. It's still in its steep phase.

---

What the MCP Adoption Timeline Means for SaaS Builders

If you're building a SaaS product in 2026, MCP is the connective tissue your AI agents run on. Ignoring it means rebuilding what the ecosystem already solved.

Here's the practical read:

• Your users expect AI agents that work with their existing tools. MCP is how you deliver that without writing 40 custom integrations.

• Your competitors are already in pilot or production. 59% of enterprise software orgs are past the evaluation stage.

• The protocol is stable. Linux Foundation governance, Apache 2.0 license, a spec that's now on version 2025-11-25. This isn't going to be deprecated in 18 months.

Three decisions the MCP adoption timeline forces right now:

01. Expose your product as an MCP server. If your SaaS isn't reachable via MCP, AI agents - including your users' own agents - can't work with it natively.

02. Build your internal agents on MCP-native infrastructure. Custom connectors are technical debt. MCP-native agents are composable by default.

03. Choose infrastructure that keeps pace with the spec. The 2025-11-25 release added async tasks, OAuth M2M, and parallel tool calls. Your agent platform needs to support these, not lag behind them. (For what's coming next, see the MCP 2026 roadmap.)

The Model Context Protocol history is still being written. But the direction is clear: MCP is the standard. The question is how fast you move on it.

---

Key Takeaways

What you need to remember:

• MCP went from a two-engineer project to a Linux Foundation standard in 17 months.

• 97M+ monthly SDK downloads - faster adoption than React.

• 41% of enterprise software orgs are in production. 30% more are in pilot.

• 70% of MCP servers come from B2B companies. This is enterprise infrastructure, not a dev toy.

• 72% of builders expect MCP usage to grow in the next 12 months.

• The Linux Foundation move makes MCP procurement-safe for enterprise buyers.

• 700% tool usage increase in documented agent-to-MCP integrations.

• The window to be an early mover is closing. The window to be a laggard is opening.

---

FAQ

What is the Model Context Protocol (MCP)?

MCP is an open standard that lets AI agents connect to external tools, data sources, and APIs through a single, consistent interface. It was created by Anthropic and is now governed by the Linux Foundation's Agentic AI Foundation.

When was MCP launched?

MCP was publicly launched on November 25, 2024, after development began in July 2024 by Anthropic engineers David Soria Parra and Justin Spahr-Summers.

Why did Anthropic donate MCP to the Linux Foundation?

To make MCP vendor-neutral and community-governed. On December 9, 2025, Anthropic donated MCP to the Agentic AI Foundation (AAIF) under the Linux Foundation, co-founded with Block and OpenAI. The move ensures no single company controls the protocol's roadmap.

How widely adopted is MCP in 2026?

Very. 41% of enterprise software organizations are in limited or broad production with MCP (Stacklok, 2026). There are 21,000–23,000+ public MCP servers, and the Python + TypeScript SDKs see 97M–110M downloads per month.

Is MCP only for Anthropic's Claude?

No. MCP is model-agnostic. OpenAI, Google DeepMind, and Microsoft all adopted MCP in early 2025. It works with any AI agent or LLM that implements the protocol.

What is the latest MCP specification version?

The current stable spec is version 2025-11-25, released on MCP's first anniversary. It adds async task workflows, OAuth machine-to-machine flows, parallel tool calls, and the .mcpb bundle format.

What does the MCP adoption timeline mean for SaaS companies?

It means the window for early-mover advantage is narrowing fast. SaaS products that expose MCP servers and build on MCP-native agent infrastructure will integrate natively with the tools and agents their users already run - without custom connectors.

---

Useful Sources

• Model Context Protocol - Official Specification (2025-11-25)
• Anthropic - MCP Launch Announcement (November 2024)
• Anthropic - Donating MCP to the Agentic AI Foundation (December 9, 2025)
• Linux Foundation - AAIF Formation Press Release (December 9, 2025)
• MCP Blog - First Anniversary & New Spec (November 25, 2025)
• MCP Blog - MCP Joins the Agentic AI Foundation (December 9, 2025)
• Stacklok - State of MCP in Software 2026 (PDF)
• Zuplo - State of MCP Report

---

Building AI agents into a SaaS product? Explore how Ginger Labs helps you ship MCP-native agents without rebuilding your infrastructure from scratch.