Standard

Model Context Protocol (MCP)

An open, vendor-neutral protocol — frequently called "**USB-C for AI**" — that standardizes how reasoning engines (LLMs and agentic runtimes) discover, invoke, and exchange context with tools and data sources. Uses **JSON-RPC 2.0** over multiple transports. The MCP architecture cleanly decouples the reasoning engine from the data systems via a three-entity model: - **MCP Host** — the runtime environment housing the LLM (e.g., Claude Desktop, agentic IDE, agent orchestrator). - **MCP Client** — the connector inside the host that negotiates the protocol handshake, retrieves context, and formats tool calls. - **MCP Server** — the standalone microservice that securely exposes specific tools, temporal memory, or S3 resources to clients.

29 connections 7 posts

Definition

What it is

An open, vendor-neutral protocol — frequently called "**USB-C for AI**" — that standardizes how reasoning engines (LLMs and agentic runtimes) discover, invoke, and exchange context with tools and data sources. Uses **JSON-RPC 2.0** over multiple transports. The MCP architecture cleanly decouples the reasoning engine from the data systems via a three-entity model: - **MCP Host** — the runtime environment housing the LLM (e.g., Claude Desktop, agentic IDE, agent orchestrator). - **MCP Client** — the connector inside the host that negotiates the protocol handshake, retrieves context, and formats tool calls. - **MCP Server** — the standalone microservice that securely exposes specific tools, temporal memory, or S3 resources to clients.

Why it exists

Pre-MCP, every agentic integration required bespoke API connectors — custom Boto3 logic, custom database adapters, custom file-read tools. MCP replaces that brittle per-integration glue with a uniform handshake that exposes capabilities at runtime, so an agent can dynamically discover what an MCP-attached server offers and call it through a standard JSON-RPC interface. The S3-relevance is direct: the AWS-published [MCP Server for AWS S3](https://aws.amazon.com/blogs/storage/implementing-conversational-ai-for-s3-tables-using-model-context-protocol-mcp/) lets an agent list buckets, read Iceberg tables via Daft, or append records to a data lake without the developer hardcoding any AWS SDK calls.

Primary use cases

Standardized tool/resource exposure to LLMs, S3-backed agent memory accessed through MCP servers, Iceberg-table-as-MCP-resource for conversational analytics, agentic IDE integrations (Claude Desktop, Cursor, VS Code, JetBrains, Xcode), distributed multi-agent systems sharing a common discoverable operational framework.

Recent developments

Latest signals

  • 97M monthly SDK downloads, 81K+ GitHub stars (March 2026). In 18 months since Anthropic open-sourced MCP (November 2024), it has become the de facto standard for AI agent integration. Supported by every major AI vendor — Anthropic, OpenAI, Google, Microsoft, AWS. Per DEV — Complete Guide to MCP in 2026.

  • 10,000+ public MCP servers across registries. The MCP server ecosystem has exploded — from PulseMCP-tracked count (was 14K mid-2025) to broader cross-registry totals exceeding 10K public servers by March 2026. Works with Claude (Anthropic), GPT/ChatGPT (OpenAI), Gemini (Google DeepMind), Copilot (Microsoft), Cursor, Replit, VS Code Copilot. Per DEV — Complete MCP Guide 2026.

  • AWS MCP Server now GA (May 2026). AWS announces general availability of the AWS MCP Server — managed server giving AI coding agents secure, auditable access to AWS services through MCP. IAM-based guardrails + CloudWatch metrics + CloudTrail logging. Per AWS What's New — AWS MCP Server GA.

  • AWS ECS deployment guide for custom MCP servers. AWS published a three-tier MCP application reference on Amazon ECS: Gradio UI + AI Agent (Bedrock-powered) + FastMCP server, with Service Connect for service-to-service comms + Express Mode for load balancing. Per AWS — Deploying MCP Servers on ECS.

  • M×N → M+N problem reframing. MCP transforms the M-clients-times-N-tools integration cost into M+N: build M clients + N servers, requiring only M+N implementations. The structural reason MCP became standard so fast. Per SSNTPL — What is MCP.

  • H2 2026 roadmap: stateless servers + automatic discovery + A2A. Stateless server operation, automatic discovery via MCP Server Cards, and agent-to-agent (A2A) coordination mature in H2 2026 — evolving MCP from single tool connections to the foundational multi-agent orchestration infrastructure. Per Ruh.AI — AI Agent Protocols 2026.

  • MCP turned object storage into an agentic tool surface (June 2026 wave). S3 is now exposed to agents as governed MCP tools — txn2/mcp-s3, MinIO's AIStor MCP Server, and the Amazon S3 Tables MCP Server let models browse, read/write, and analyze objects in natural language, making storage a machine-operated memory-as-a-service plane. Per MinIO — Model Context Protocol (AIStor) and GitHub — txn2/mcp-s3.

Connections 29

Outbound 5
scoped_to2
AI Runtime InfrastructureS3
enables1
AI Runtime Infrastructure
used_by2
Mem0Vestige
Inbound 24click to expand
extends4
txn2/mcp-s3AIStor MCP ServerS3 Tables MCP ServerMCP Tasks Primitive (SEP-1686)
implements4
VestigeOpenMemory MCPAmazon Bedrock AgentCore RuntimeMCP Knowledge Graph
orchestrates1
LangGraph
integrates_with1
Letta
scoped_to6
OWASP MCP Top 10MCP Tasks Primitive (SEP-1686)MCP Knowledge GraphContext Injection & Over-Sharing (MCP10)Confused Deputy Problem (MCP)Tool Discovery Governance Gap
governs2
OWASP MCP Top 10MCP Gateway
complements2
Agent2Agent (A2A) ProtocolAgent Communication Protocol (ACP)
depends_on1
FAME Architecture
augments2
Direct Corpus Interaction (DCI)Catalog-Centric Control Plane
constrained_by1
Tool Discovery Governance Gap

Featured in

2026-07-05 The Frontier Moved Again — and the Floor Compounded In June we argued the AI stack had split into a closed, expensive frontier and an open, cheap floor. Six weeks later the frontier proved the point the hard way: Claude Fable 5 was pulled offline for 19 days by an export-control order, then restored with a retention mandate that quietly bars the most sensitive data from ever touching it. Meanwhile the floor didn't just hold — it compounded. Object storage got RDMA-fast, the query engines converged on one substrate, and the data plane grew a control plane built for agents instead of humans. This is the field report. 2026-06-27 The Storage Cost Inversion: When Object Storage Grew a Brain A 2026 NAND/flash shortage and a wave of cloud storage price hikes made fast storage scarce — and pushed AI memory onto S3. But object storage didn't just become the cheap tier. Under pressure it became the active substrate: an agentic data plane, an RL training buffer, and a hot KV-cache memory pool. 2026-06-10 The Catalog Became the Control Plane — for Humans and Agents Alike Within one quarter, Apache Polaris, Unity Catalog, and Apache Gravitino all converged on the same role: the catalog is no longer where you look up where a table lives — it's where access is granted, credentials are minted, query planning happens, and AI agents are let in or kept out. The companion shift to 'the catalog is becoming a database': the catalog is also becoming the control plane. 2026-06-06 Storage Grew an Agent Interface The S3 API made storage programmable for applications. In a six-week stretch of 2026, the storage layer grew a second interface — for agents. Weaviate, Pinecone, and RustFS all shipped built-in MCP servers; cheap inference and agentic retrieval made it inevitable; and the Apache Polaris CVE cluster showed the security bill that comes due. This is the discoverability inflection for self-hosted AI data infrastructure. 2026-06-05 When Inference Became Cheaper Than Storage: The May 2026 Cost Inversion In May 2026 DeepSeek made its 75% price cut permanent. V4-Flash output is now 107x cheaper than GPT-5.5. For the first time the cost of intelligence dropped below the cost of moving data — and every storage decision made for the 2024 cost curve is now suspect. This is the continuation of the I/O-stack rewrite. 2026-06-05 The Catalog Is Becoming a Database: Metadata's Independence Day In 2024 the catalog was a metadata appendix bolted to the table format. In 2026 three signals — TreeCat's standalone engine, StarTree's sub-second serving on Iceberg, and Weaviate's MCP server — show the catalog becoming an independent database with its own query language, concurrency model, and storage layout. This changes how lakehouses are built. 2026-05-16 When AI Memory Became an Architecture: KV-Cache Persistence, MCP, and the Night S3 Got Its Memory Tier Stateful agents needed stateful storage. Tonight 33 new nodes joined the index: Mem0, Zep, LMCache, SGLang, Mooncake, MCP, cuObject, BlueField-4, Animesis CMA. The architectural thesis: AI memory infrastructure crystallized around object storage in 2026, and the persistence layer the industry settled on is S3.