Model Context Protocol (MCP)
Anthropic-created open standard for connecting AI agents to tools and data; the de facto integration layer of the agent era, now stateless after a breaking 2026 spec revision.
The Model Context Protocol standardizes how an AI agent connects to external tools and data sources. An MCP server exposes tools, resources and prompts; an MCP client inside an agent (Claude, ChatGPT, Gemini, Cursor, Codex and nearly every other agent) discovers and calls them. Anthropic published it in November 2024; within a year it was the de facto standard, and in December 2025 Anthropic donated it to the Linux Foundation’s Agentic AI Foundation, which also hosts A2A.
The 2026-07-28 breaking change
Spec revision 2026-07-28 removes protocol-level sessions, the Mcp-Session-Id header and the initialize handshake, making MCP stateless: protocol version and capabilities now travel with each request in _meta. Anything built against 2025-11-25 or earlier needs migration.
Where MCP shows up in this wiki
MCP is the connective tissue between almost everything here: Figma’s 20+ MCP integrations, Pigment’s server for Claude, ChatGPT, Cursor and Codex, Pliant’s agent payments, Claude Security’s connector, Tencent’s search API, REA (reverse-engineering tools for agents), Respan’s gateway and the Gemini Enterprise agent. IBM watsonx Orchestrate discovers and evaluates MCP servers across clouds. Google says its Universal Commerce Protocol is MCP-compatible.
Assurance angle
MCP availability does not equal enterprise readiness. Most public servers weren’t built with authentication, RBAC or audit trails in mind, which is exactly the gap a customer assurance review should probe. MCP servers also hold credentials: after the GitBook attack, GitBook’s MCP users had to re-authenticate. In November 2026 German host Mittwald said an attacker it believes used AI tools got in through its own MCP server for customer AI assistants and compromised 56 customer projects (Mittwald MCP server breach): the integration layer itself became the entry point. Cross App Access adds enterprise identity-policy routing as an authorization extension.
Timeline
Every dated event in the wiki that involves Model Context Protocol (MCP), newest first.
At Adobe MAX, Adobe released a local MCP exposing endpoints across Illustrator, Photoshop, After Effects and Premiere so chatbots such as ChatGPT, Claude or Gemini can use them, plus an always-on agent and new AI features in Photoshop, Premiere and Firefly.
German web host Mittwald says an attacker exploited a flaw in its MCP server for the mStudio platform and went undetected for eight days, compromising 56 customer projects. It believes AI tools were used because steps came seconds apart; it gave no other evidence (company assessment).
- Mistral adds managed agents and apps to Studio
Managed agents with skills, sandboxes, subagents and hooks, event-triggered workflows, observability and evals, and a Mistral MCP server for one-command deploys; paired with Mistral Large 4.
- Pliant raises $80M
Funding: oversubscribed Series C for B2B payments, led by Seaya Nova and Speedinvest; lets outside AI agents connect via MCP with permissions, controls and audit trails.
Proactive agent app for Pro/Max users; Claude Security public beta extends to Claude Enterprise with an API and MCP connector.
Sources: Anthropic, “Introducing Claude Duet”
- Microsoft patches a Copilot flaw that let it rewrite its own MCP config (Visual Studio Copilot CVE-2026-103572)
A trailing dot in a file name slipped past Copilot's path check in Visual Studio and SQL Server Management Studio, letting the agent edit its own MCP and settings files and run commands, including via Prompt injection. Microsoft rated CVE-2026-103572 an Important RCE and fixed it on Patch Tuesday but paid no bounty, the researcher says.
Sources: Embrace The Red, “Visual Studio Copilot and SSMS Remote Code Execution (CVE-2026-103572)”
- Gemini Enterprise gets a unified agent
The agent gets its own Workspace account and email, works with MCP servers and can use Claude.
- A2A joins the Agentic AI Foundation
A2A, which Google launched in April 2025 and donated to the Linux Foundation, joins MCP as a hosted AAIF project.
Sources: Latent Space, “AAIF”; Techstrong.ai, “Google Moves A2A Under Agentic AI Foundation”
- MCP spec revision 2026-07-28 makes the protocol stateless
Breaking change: removes sessions, the Mcp-Session-Id header and the initialize handshake.
Sources: modelcontextprotocol.io, “Key Changes (specification 2026-07-28)”
Connections
Model Context Protocol (MCP) and everything one link away. Drag to move, scroll to zoom, click a node to open it.Open in the full graph →
What links here
Protocol Agent Client Protocol (ACP) · Agent2Agent (A2A) · Cross App Access (XAA) · Google Universal Cart (Universal Commerce Protocol) · Personal Agent Protocol
Organization Agentic AI Foundation (AAIF) · Linux Foundation
Concept AI agent
AI lab Alibaba (Qwen) · Anthropic · Google DeepMind · Mistral AI
Product Citrix SecurSpaces · Claude Agent SDK · Claude Code · Claude Security · IBM watsonx Orchestrate · REA (Reverse Engineer Anything) · Sysdig AI Defense · Tencent Yuanbao Search API · Zapier and Make
Company Composio · Figma · Mittwald · Okta · Pigment · Pliant · Respan
Incident GitBook AI-agent attack · Mittwald MCP server breach · Visual Studio Copilot CVE-2026-103572
Model family Qwen
Publication Tracking AI news