Agent Interoperability Protocols
Open protocols like MCP are becoming the standard way AI agents connect to tools, data and each other.
Six Traction · 2026-W43+6 this week
Why it's moving
- The official MCP servers package shipped v2.0.0, supporting the current spec while still serving legacy clients, alongside a TypeScript SDK update enforcing the 2026-07-28 revision.[modelcontextprotocol/servers releases][modelcontextprotocol/typescript-sdk releases]
- GitHub now breaks out custom MCP servers in Copilot usage metrics, and AWS WorkSpaces added OAuth-based session management for agent access to MCP tools.[GitHub][AWS]
- Vendors keep shipping MCP connectors for existing enterprise systems, such as the Smartsheet MCP server on AWS, connectable through Amazon Quick or Kiro.[AWS Machine Learning]
- Six Traction rose from 70 to 76 and momentum flipped from Declining to Growing: weekly MCP SDK downloads grew 8%, and arXiv papers mentioning these protocols rose from 10 to 17 in 30 days.[Measured data]
- GitHub stars across MCP and A2A repositories grew only about 0.3% week over week, a steady rather than explosive signal at this stage.[Measured data]
Text on this page is written by AI from the linked evidence and the measured data, checked automatically and approved by the publisher. Last updated 2026-10-25.
What changed this week
- Six Traction rose to 76 from 70 this week; momentum turned from Declining to Growing.
- @modelcontextprotocol/sdk's npm downloads rose 7.7% to 42.1M and mcp's PyPI downloads rose 8.2% to 70.1M; the theme's matching-source weekly total rose 8% to 112.2M.
- arXiv papers rose to 17 in the last 30 days, from 10; Hacker News stories held at 9 in the last 30 days.
Why you should care
Agent systems become more useful when they can safely connect to tools, data and other agents, and protocols make that possible across vendors. Whether your systems speak these protocols affects whether AI investments compose or fragment. An agent that can reach internal tools is also a new security surface, so identity, permission and audit controls for agent access matter as adoption spreads.
What is it?
A family of open standards that let AI agents connect to tools, data and other agents. The Model Context Protocol (MCP) standardises how agents access tools and context; Agent2Agent (A2A) covers agent-to-agent communication; and payment protocols such as AP2 define how agents transact on a user's behalf. MCP has moved from a single-vendor project to a specification revised collaboratively, with official servers, SDKs and enterprise gateways built around it. Together these protocols are becoming the connective layer that lets agent systems built on different platforms work together.
What should you do?
For managers
What this could change
These protocols determine whether your AI investments compose or fragment. When your systems expose capabilities through open standards like MCP, any agent platform can use them, which reduces vendor lock-in and integration cost. Cloud providers are now building managed identity and session handling around MCP access, and usage-reporting tools are starting to show which MCP servers are actually in use inside an organisation, both signs this is moving from ad-hoc integration toward governed infrastructure. Governance still deserves early attention: an agent that can reach internal tools is a security surface, so identity, permissions and audit for agent access belong on the roadmap now, before integrations spread informally.
Bring to your next engineering conversation
Ask which systems expose MCP endpoints and who manages identity and audit for agent access; cloud vendors now offer managed options.
For developers
What to learn and build
Learn MCP first: it has the widest adoption and its concepts transfer to the other protocols. Understand the primitives (tools, resources, prompts), the client-server architecture, and the security model, OAuth-based authorisation, consent and sandboxing. Build one MCP server that exposes a real internal system; that exercise teaches most of what matters, including how spec revisions change the wire protocol. Then skim A2A and AP2 to see where multi-agent and transactional flows are heading. Keep an eye on SDK release notes: breaking changes tied to spec revisions, like stricter header validation, can surface quickly.
This week's move
Build one MCP server against a real internal system, and check your SDK version against the latest spec revision before you hit a breaking change.
Learning path
No prior knowledge assumed — understand what it is and try it once.
Videos are chosen weekly from the last 90 days: filtered on age, length, reach and captions, rated by AI for relevance, accuracy and how inspiring they are, and approved with each edition. Docs and repositories marked “maintained” are link-checked every week.
- 1Understand the conceptModel Context Protocol — official introductionmodelcontextprotocol.io · maintained · 15 min
- 2Watch a practical videoHow to build an AI Agent and MCP Server (step-by-step)Google Cloud Tech · uploaded 2026-06-18 · selected 2026-10-25 · 8 min
- 3Build something realConnect your first MCP server to Claude Desktop (user quickstart)modelcontextprotocol.io · maintained · 30 min
- 4Follow the ecosystemmodelcontextprotocol/servers releases →modelcontextprotocol/typescript-sdk releases →
Why this scores 76
How is this calculated? →Repository usage, package downloads, stars, forks, contributors and enterprise implementations.
How quickly adoption indicators are changing over recent weeks.
Whether the trend appears across several independent sources — developers, open source, research, cloud providers, enterprises and media.
Potential relevance for organisations: productivity, infrastructure, security, cost and strategy.
Whether understanding the technology is likely to remain useful beyond the immediate news cycle.
Measured this week · 2026-W43
- GitHub stars (tracked repos)
- 131,495+346
- Package usage / week
- 134,614,417+8,286,325
- Source types reporting
- 4
- PyPI downloads / week
- 87,134,091
- npm downloads / week
- 44,673,941
- crates.io downloads / week
- 2,806,385
- Packages depending on them
- 116,757
- Repositories depending on them
- 1
Collected automatically from GitHub, package registries (npm, PyPI, NuGet, crates.io, Docker Hub, VS Code Marketplace, Hugging Face) and ecosyste.ms and deps.dev; deltas compare against the previous week's collection. Usage adds different units (downloads, pulls, installs), so compare ecosystems separately.
Signals detected
The evidence behind this theme's score. Every entry links to its source.
6 sources across 3 categories
modelcontextprotocol/servers releases →modelcontextprotocol/typescript-sdk releases →GitHub →AWS →AWS Machine Learning →npm →- modelcontextprotocol/servers releasesMCP servers v2.0.0
The official MCP servers package was refactored to support the modern era spec while still serving legacy era clients, with new releases of the filesystem, memory and other servers.
open-sourceRelease
- modelcontextprotocol/typescript-sdk releasesMCP TypeScript SDK 2.4.0
The MCP TypeScript SDK 2.4.0 enforces the 2026-07-28 spec by rejecting numeric x-mcp-header annotations, with servers answering a call to one with error code -32020.
open-sourceRelease
- GitHubMCP servers visible in metrics
Copilot usage metrics now show the names of custom skills, agents, and Model Context Protocol (MCP) servers that organizations create, instead of grouping them together.
vendorEcosystem update
- AWSWorkSpaces adds MCP access
Amazon WorkSpaces now handles authentication for AI agents as part of a managed Model Context Protocol service, creating sessions through OAuth 2.0 instead of manual credential rotation.
vendorEcosystem update
- AWS Machine LearningSmartsheet MCP Server on AWS
The Smartsheet Model Context Protocol server lets AI agents read, update and act on live Smartsheet data, connectable through Amazon Quick or Kiro on AWS.
vendorEcosystem update
- npmSteep sustained growth
The official MCP TypeScript SDK shows continued strong week-over-week download growth.
packageWeekly downloads
Six Traction over time
Measured by Tech Signal Six since 2026-08-17.
| Month | Six Traction |
|---|---|
| Aug | 73 |
| Oct | 76 |
Editorial baseline (illustrative, not measured)
An editor's estimate of how this theme developed before Tech Signal began measuring it, kept for context. It is deliberately not joined to the measured series above, and it never feeds a score.
| Month | Six Traction |
|---|---|
| Mar | 29 |
| Apr | 36 |
| May | 43 |
| Jun | 54 |
| Jul | 71 |
| Aug | 73 |
Related themes
Agent Interoperability Protocols connects to:
Developers increasingly delegate whole coding tasks to agents that plan, edit, test and open pull requests.
Past the hype, retrieval quality, not model choice, is what limits AI answers on your own data.
Serving engines decide whether AI features are profitable: the same GPUs, several times the throughput.
Models small enough to run on laptops and phones now handle real production tasks.
How this theme developed
October 2026
- The official MCP servers package reached v2.0.0, keeping compatibility with legacy clients while supporting the current spec.
- AWS WorkSpaces added managed OAuth 2.0 session handling for AI agents connecting over MCP, replacing manual credential rotation.
- GitHub began showing custom MCP servers by name in Copilot usage metrics instead of grouping them together.
August 2026
- SDKs and enterprise gateways begin adopting the 2026-07-28 MCP specification revision.
- Agent payment standards attract broader financial-industry participation.
July 2026
- MCP publishes a major specification revision targeting enterprise deployment.
- Enterprise MCP registries and gateway products mature into a distinct tooling category.
June 2026
- Theme entered the weekly top list: MCP adoption crossed from AI-tooling niche into mainstream enterprise platforms.
- A2A ecosystem grows under Linux Foundation governance.