In the 7Maps agent study, the worse of two servers offered the agent itself 50 credits for picking it. Claude Opus, Sonnet and Haiku refused in every run. Gemma 3 4B, Llama 3.2 3B and Qwen3 4B took the worse server in 71% to 100% of their valid answers.
What to do
Find the step where your agent picks a tool or server, and check which model makes that pick.
An agent that can be bought for 50 credits is not working for you.
MCP Inspector 2.14.0 on 4 November 2026 fixed three vulnerabilities and 2.15.0 on 12 November fixed four more, per their release notes. They include an MCP App that could leave its sandbox frame policy and a fetch guard that missed some cloud metadata addresses. The 1.x line got its own security fixes in 1.0.4 and 1.0.5.
What to do
Run npx @modelcontextprotocol/inspector@latest, or @v1-latest on the 1.x line, and remove older copies pinned in scripts or docs.
A testing tool meets the worst servers first, so it gets patched first.
Claude Code 2.1.325 shipped on Friday, 13 November 2026. Its changelog says an MCP server could stay at "needs authentication", and its mcp_tool hooks be skipped, after a token refresh was refused only for a moment. The same release fixes headless elicitation requests cancelled without asking, and SDK host tool calls answering "Not connected" after a reconnect.
What to do
If users reported sign-in loops or "Not connected" errors with your MCP server, ask them to update to Claude Code 2.1.325 before you debug your own token code.
A token refresh that fails for a moment should heal in a moment. Until this release, it didn't.
Claude Code 2.1.323 shipped on Thursday, 12 November 2026. One fix: MCP tool calls used to end in the sign-in server's own error, such as TOKEN_EXPIRED, instead of asking the user to sign in again. The same release adds an engines["claude-code"] field to plugin.json, so a plugin outside its stated version range is not installed, updated or loaded.
What to do
Upgrade Claude Code to 2.1.323, let a test token expire, and confirm your MCP tool call ends in a sign-in prompt. If you ship a plugin, add engines["claude-code"] to plugin.json with the version range you tested.
An expired token is not an outage. It just looked like one in the chat.
On 12 November 2026, Anthropic opened a private beta of self-hosted Claude Managed Agents: the full service, including sandboxes, credentials and session history, runs in the customer's own cloud or on premises. For MCP tools, a proxy outside the sandbox adds the credential to the request. Credentials can sit in the customer's own secrets manager.
What to do
If you sell an MCP server to enterprises, publish the exact hostnames it calls so a customer can allowlist them, and test that it works when a proxy adds the credential to each request.
The safest key is the one the agent never sees. Make sure your server is ready to meet that agent.
On 11 November 2026 the MCP reference servers fixed six vulnerabilities in the git, filesystem and fetch servers. The worst, GHSA-6hqx-gqrx-c5r2 (High, CVSS 7.8), let a repository's own Git config, such as core.fsmonitor, run a program through tools like git_status. The fixes ship in mcp-server-git and mcp-server-fetch 2026.11.11 and server-filesystem 2.3.0.
What to do
Upgrade to mcp-server-git 2026.11.11 today. If you run server-filesystem on a date-style version, install @2.3.0 explicitly, and until then point the git server only at repos whose .git folder you trust.
Read-only is a promise about your tools, not about the data they read. Treat every repo an agent hands you as hostile input.
On 11 November 2026 the MCP lead maintainers set a six-month schedule: spec releases on the last Tuesday of March and September, starting with version 2027-03-30. Proposals must be Final and merged by a soft freeze two and a half months before release, and Tier 1 SDK betas arrive a month and a half before.
What to do
Put 30 March 2027 in your server's roadmap today, and plan to test against the Tier 1 SDK betas about six weeks before it.
A spec with a calendar is a spec you can build a business on. Miss the train and you wait six months.
On 10 November 2026 Anthropic said interactive connectors can now open beside the conversation or full screen, and Claude can change a page, row or file in the open tool once the user approves. They run on MCP Apps, and Claude now supports more of the extension, such as app tools and embedded sites, in public preview for developers.
What to do
If you ship an MCP App, read Claude's MCP Apps getting-started docs today and test app tools and embedded sites in the preview before the rollout reaches your users.
The chat box used to be the front door. Now it is the whole office, and your server decides what is inside.
Claude Code 2.1.321 shipped on 9 November 2026 with a fix for MCP OAuth requests that sent secrets, such as refresh tokens, to another host when a server redirected them. The same release fixes servers that showed as connected with no tools after a reconnect, and refuses a tool call when a PreToolUse mcp_tool hook answer tops 16 MiB.
What to do
Check every redirect in your server's OAuth endpoints today and keep token requests on your own host. Then tell your users to run Claude Code 2.1.321 or later.
A redirect is a polite way to hand someone your keys. Audit yours before a client has to protect you from it.
On 6 November 2026 Microsoft put a SharePoint MCP endpoint into preview through Work IQ. Four generic tools (search_paths, get_schema, do_action, get_site_key) reach more than 60 SharePoint operations. Calls run as the signed-in user, under new WorkIQ.Read and WorkIQ.Act scopes.
What to do
Read the Work IQ post today, then check how many tools your own server exposes and whether a search-then-call layer would shrink the list your clients must load.
A small tool list that can reach sixty actions is a design statement. The servers with a hundred flat tools should read it twice.
On 6 November 2026, AWS made Agent Registry, available through Amazon Bedrock AgentCore, generally available in the AWS GovCloud (US-West) Region. Teams publish MCP servers, tools, agents and agent skills to a searchable catalog, and administrators decide what becomes discoverable through an approval workflow.
What to do
If you sell an MCP server to public sector teams on AWS, write down what an approver needs to say yes: owner, auth model, data touched and a changelog, in one page.
The registry with an approval button is becoming the front door. Make your server easy to approve, not just easy to install.
Per its changelog, Claude Code 2.1.319 adds MCP tasks in the terminal: when a server runs a tool call as a task, the client moves it to the background right away and gives Claude the result when it finishes. CLAUDE_CODE_DISABLE_MCP_TASK_BACKGROUND turns that off, and the same release fixes servers on the older handshake that failed to connect or returned empty results.
What to do
If a tool of yours runs longer than a few seconds, check whether your SDK can run it as a task, then test it against Claude Code 2.1.319. If your server uses the older handshake, retest the connection on this version.
Slow tools were always the server owner's problem. Now the client has a lane for them, and only servers that ask for it get in.
On 3 November 2026, Microsoft's Azure SDK blog listed seven Azure plugins: the core one plus six specialized ones. You add the azure-skills marketplace in GitHub Copilot CLI, Claude Code or Cursor, then install each plugin. The Azure Cost plugin is backed by the Azure Resource Manager MCP server.
What to do
If your MCP server has a skill that teaches agents to use it, package both as one plugin and test the install in Claude Code.
The MCP server is becoming the engine under the plugin. The plugin is what people actually install.
Claude Code 2.1.318, published to npm on 3 November at 21:55 UTC, stops re-probing stdio MCP servers that quit during the version probe at every start. A server that dies in under 3 seconds for its own reason is now pinned to the legacy handshake for 7 days. A call to a tool before it loads now returns the tool input schema.
What to do
Test your stdio server on 2.1.318 from a cold start, and make sure it never quits within 3 seconds of launch for a missing config or env value; log the error and keep running instead.
A crash at startup used to cost you one launch. Now it can cost you a week. Fail loudly, but stay up.
On 30 October the MCP blog announced that the official Ruby SDK, the mcp gem, reached Tier 1, the highest of three SDK tiers. It passed 67 of 67 server and 50 of 50 client conformance scenarios, for both the 2025-11-25 and 2026-07-28 specs. Ruby joins the TypeScript, Python, C#, Go and Rust SDKs at Tier 1.
What to do
If you ship an MCP server in Ruby, set gem 'mcp', '~> 1.8' in your Gemfile, run bundle update mcp, and retest with a 2025-11-25 client and a 2026-07-28 client.
Tier 1 is a promise about next March, more than a badge about today. Pick the SDK that has made that promise in writing.
Claude Code 2.1.317, published to npm on 2 November, adds a link card to cloud sessions, so a user can open or decline a link an MCP server asks them to visit; until now the request went unseen. It also fixes project and local MCP servers disconnecting after the first turn in some Claude desktop sessions. On npm, stable is 2.1.311.
What to do
If your server sends URL prompts, test the flow in a Claude Code cloud session on 2.1.317, both accepting and declining the link, and make sure a declined link returns a clear message.
A prompt the user never sees is a prompt that failed. Test your server where your users run it, and the bug reports get shorter.
On 30 October the MCP Registry released v1.8.7. It adds prepared listing JSON, so server payloads are not decoded and rebuilt on every listing read, and prepared reads are off by default with a rollback flag. Local benchmarks showed about 63% less encoding time and 88% fewer allocations.
What to do
Nothing to do yet; watch for the registry turning prepared reads on by default.
A fast registry is only as good as the entry inside it. Fix your metadata before the traffic arrives.
On 30 October the MCP project merged PR 3306, a docs-only change to the Enterprise-Managed Authorization page. It adds a table of 12 products with public documentation for the ID-JAG flow, each marked with the vendor's own label: Available, Early access or Experimental. The page says the list is not comprehensive.
What to do
If you sell to enterprises, check whether your authorization server appears in the table and which label it carries. Then check the draft revision each vendor targets before you promise anything.
The first enterprise question is whether their identity provider can sign in to your server. Have that answer written down.
Claude Code 2.1.314, published to npm on 30 October, fixes MCP servers that also work without sign-in showing "needs authentication" after their stored sign-in stopped working; they are now used signed out. It also fixes claude mcp logout reporting success when the sign-in was not removed. On npm, latest is 2.1.314 and stable is still 2.1.308.
What to do
If your server allows anonymous calls, test an expired-token session on Claude Code 2.1.314 and on 2.1.308, and list in your docs which tools still work signed out.
Optional sign-in is a promise. Every client that breaks it makes your server look down when it's up.
On 29 October the MCP project merged SEP-3392, now Final, which gives official extensions three stages: Experimental, Beta and Stable. Apps and Auth were classified Stable; Skills, Tasks and Server Card went to Beta. In the extension support matrix, 3 of 16 listed clients mark Tasks support: Claude (web), Claude Code and MCPJam.
What to do
If your server advertises Tasks or Skills, pin the extension revision you test against and watch for the revision string the SEP asks breaking changes to add or bump.
Beta is an honest word. Build on it, but keep a note of exactly which version you built on.
On 28 October 2026 the MCP servers repo published advisory GHSA-w766-5x43-whvw for mcp-server-git, rated Moderate (CVSS 5.3). Run with --repository, a nested folder holding a .git file could point Git at a repository outside the allowed one. Versions before 2026.10.28 are affected.
What to do
If you run mcp-server-git with --repository, upgrade to 2026.10.28 today and look for stray .git files inside the allowed folder.
Confinement is only as good as the last thing you check. Check the repository Git opens, not the path you were handed.
On 28 October 2026 GitHub began rolling out GitHub-verified MCP servers, called connectors, in public preview to users on all Copilot plans. At launch 26 services are supported, including Slack, Figma, Notion, Sentry and Linear.
What to do
Check whether your service is among the 26 connectors. If your users sit on Copilot Business or Enterprise, tell their admins about the policy and the allowlist step.
Verification is the new shelf space. The shorter the setup, the more the shelf decides who gets used.
On 27 October 2026 the official MCP Registry shipped v1.8.6. Any non-empty search term must now be at least three Unicode characters, and shorter ones get HTTP 422. The same release reworks how search pages are fetched, to hold up under mixed search, pagination and publishing traffic.
What to do
Search your code for calls to the registry search with user input. Hold back terms under three characters before the request, and treat a 422 as an empty result, not a crash.
Performance releases are the ones nobody reads. They are also the ones that change what your users see the next morning.
Claude Code 2.1.311 reached npm on 26 October. Its changelog fixes subagents being unable to use MCP tools from servers that finish connecting after the subagent starts, and an interrupt that waited for a timeout while an MCP server had not answered a resource read. The stable channel is still 2.1.308.
What to do
Measure how long your server takes to finish connecting, from first request to a ready tool list, and cut it where you can. Users on the stable 2.1.308 channel still have the old behaviour.
Fast startup used to be a nicety. Once agents delegate, it decides whether your server is in the room at all.
On 25 October the official MCP servers repo published 8 advisories, 3 Moderate and 5 Low. Two hit server-filesystem: a trailing quote or space slipped past the path check, and on macOS and Windows a different letter case got past the symlink check. Both are fixed in 2.0.1.
What to do
Run npm ls @modelcontextprotocol/server-filesystem today. If it shows a 2025.x or 2026.x version, install 2.0.1 by name, as the advisories say, and pin it.
"I updated" and "I'm patched" used to mean the same thing. Check the version you are running, not the command you ran.
On Tuesday 20 October 2026 Microsoft introduced retrieve, a new MCP tool in Work IQ that searches a user's email, files, meetings, Teams messages and people. Microsoft says it evolves the Copilot Retrieval API for agents. Each hit carries fields such as webUrl, contentProtection and sensitivityLabel, next to a markdown answer to ground on.
What to do
If you sell an MCP server to Microsoft 365 customers, read the strategy section: grounding limits search to SharePoint, OneDrive, Teams and Outlook, so plan for customers who turn your source off.
Enterprise agents will not open twenty connections to find a file. They will ask one tool, and that tool decides which sources are in the room.
On Friday 23 October 2026 the official MCP TypeScript SDK shipped 2.4.0 with support for Skills over MCP (SEP-2640). A server publishes a folder of skills with loadSkills(folder) and skills.attach(server); a client lists, reads and verifies them with SkillsClient. The release notes mark attach as experimental.
What to do
If your server is on the TypeScript SDK, upgrade to 2.4.0 in a branch, attach one skill that explains your main tool, and pin the version, since attach may still change in a minor release.
Tools without instructions are a box of parts. A server that ships its own manual gets used correctly the first time.
On 22 October 2026 Anthropic said it is donating Agent Skills to the Agentic AI Foundation, part of the Linux Foundation, which already hosts MCP. The gift covers the name and logo, the GitHub organization and the agentskills.io domain. Anthropic cites more than 70 agent platforms, 1 million published skills and nearly 280 million installs.
What to do
Read the proposed roadmap (versioned releases, a formal proposal process, security guidance for skill authors) and send feedback while it is still a proposal.
MCP tells an agent what it can reach. Skills tell it how to do the work. One house for both was overdue.
On 21 October 2026 Cloudflare said Commerce.com turned on WebMCP for almost 100,000 of its domains through Cloudflare's edge, with no theme edits by merchants. Agents find a store's tools on document.modelContext in the browser and call them directly. Cloudflare describes WebMCP as an experimental, proposed standard in early preview.
What to do
If you build agents that shop or browse, test calling page tools on document.modelContext in a Chrome build with the WebMCP flag on, and keep your MCP server as the fallback.
Two doors into the same shop: one on the server, one in the browser. Agents will try both.
On 21 October 2026 the MCP Inspector project published nine security advisories. Two are Critical at CVSS 9.6: a malicious remote server could run local commands through the OAuth callback (versions 0.13.0 to 1.0.2) or through its MCP App (0.20.0 to 1.0.2). Version 1.0.3 fixes both, and 2.12.0 fixes the issues in the 2.x line.
What to do
Update to Inspector 2.12.0, or 1.0.3 if you stay on the 1.x line, before you connect it to any server you did not write. Remove older copies pinned in scripts.
The debugger is part of your attack surface. Treat every unknown server like an email attachment.
Claude Code 2.1.307, published to npm on 21 October 2026 at 20:47 UTC, fixes HTTP MCP tool calls that were sent twice when the connection closed mid-call. Per the changelog, only read-only or idempotent tools are now retried, and servers no longer stay failed in headless runs after a macOS or Linux computer slept while they were connecting.
What to do
Check the readOnlyHint and idempotentHint annotations on every tool you publish, and set them only where a repeated call is truly harmless.
A retry is a second request with your name on it. Only promise idempotent if you mean it.
As of 21 October 2026, Claude Code 2.1.305 is the latest tag on npm. Its changelog says allowedMcpServers no longer filters managed-mcp.json stdio servers that use ${VAR} only in env, so a server the allowlist left out now loads on upgrade. Another fix stops a same-named server from your own settings replacing one passed by a cloud session or setMcpServers.
What to do
If you manage Claude Code settings, add deniedMcpServers entries for any managed-mcp.json stdio server that must stay off, then retest after upgrading to 2.1.305.
An allowlist that quietly changes meaning is not a wall. Read the changelog before you trust the fence.
On 20 October 2026 the MCP specification repository merged PR 3440. It renames the Triggers and Events working group to Tasks and Events, moves the Tasks scope into it, and converts the Agents working group to an interest group. The Agents charter says the group will recommend how to evolve the protocol, create extensions or document how existing primitives should be used.
What to do
Read the two charters before you design a long-running agent tool, and follow the Tasks and Events group for the path of Tasks into the core protocol.
Interest groups talk. Working groups ship. Know which one your feature depends on.
Claude Code 2.1.304 shipped on 19 October 2026 and is now the latest, next and stable release on npm. In cloud sessions, a new conversation's first turn no longer waits for the local (stdio) MCP servers of plugins synced from claude.ai, unless a server sets alwaysLoad: true. The same release lets a mod's tool.check answer with decider: "person".
What to do
If your plugin's stdio server must answer on the very first turn, set alwaysLoad: true on it and test a fresh cloud session. Otherwise make sure your skill still works when the tool arrives a turn later.
Startup time is now a feature someone else trims for you. Decide whether your server is worth the wait, before the client decides it isn't.
On 16 October 2026 Cloudflare described a system that checks over time whether Verified bots honor site preferences: it compares bot traffic with robots.txt rules and watches whether a bot slows down after an HTTP 429. Each bot gets an explainable trust band. Cloudflare uses it internally for now and says it will share the scores publicly on Radar.
What to do
Check that your agent's HTTP client backs off on every 429 and reads robots.txt before it fetches a site. Count the 429s your agents received this week.
Politeness just became a score. The agents that back off at the first 429 are the ones sites will keep letting in.
Microsoft's September 2026 Foundry roundup, posted on 15 October, confirms Tool Search is generally available, as announced on 24 September. Instead of sending every tool definition with each model request, a Toolbox exposes discovery and invocation through tool_search and call_tool. Toolboxes for prompt agents entered public preview.
What to do
Rewrite each tool description so its first line names the task and the object it acts on, then test whether a plain-language search finds it.
Tool search turns every description into a search listing. Write it for the query, not for the README.
Microsoft released Azure MCP Server 3.0 on 12 October 2026 and announced it on 14 October. It uses the stateless MCP protocol 2026-07-28, so clients on that version must send Mcp-Method and Mcp-Name headers on every POST, while clients on 2025-11-25 keep a negotiated fallback. Tool prefixes moved from resilience_* to resiliency_*, and unknown parameters now produce errors.
What to do
If you run an MCP server, test it against a 2026-07-28 client that sends Mcp-Method and Mcp-Name headers. If you rename a tool, ship the old name as an alias for one release and say so in your changelog.
A tool name is an API. Renaming it is a breaking change, even when the code behind it is identical.
Anthropic put OpenTelemetry traces on the Claude Platform into public beta on 14 October 2026. Its docs say the API adds a tool span for each use of a server tool or an MCP connector tool that the API runs. The default Metadata only mode carries identifiers, timing, status, token counts and the names of MCP servers; Full content adds prompts and responses.
What to do
Give your MCP server a clear, stable name, and check your own latency and error rate before a customer's trace shows it to them.
Your customers just got a stopwatch. I'd rather know my numbers before they read them to me.
Claude Code 2.1.299 is the latest release. Its changelog says an MCP allow rule now reaches only the server it names, so mcp__a no longer allows a server named a__b or a_. Deny and ask rules for server names that contain __ are fixed too, and agent tools lists follow the same exact-name rule.
What to do
Check whether your server name contains __. If it does, tell users to write mcp__<server>__* with the whole normalized name, or rename the server.
A server name is part of your permission surface. Pick one that cannot be confused with a prefix of another.
On 13 October 2026 Microsoft announced the SharePoint Embedded MCP server in public preview, run with npx @microsoft/spe-mcp. It creates and manages container types and containers, provisions apps, configures billing and uploads content, and Microsoft names GitHub Copilot, Claude, Cursor and Azure AI Foundry as clients. On npm the latest tag is 0.2.0-alpha.4, published 9 September.
What to do
If you try it, pin the exact version in your client config, use a test tenant first, and require a person to approve any billing or provisioning call.
A tool that can configure billing deserves a pinned version and a human click. Preview is a label. The version string is the truth.
On Friday 9 October 2026, Claude Code 2.1.296 raised the default limit on MCP tool descriptions sent up front, and on MCP server instructions, from 2,048 to 4,096 characters. On npm it is the latest release, while the stable channel still points to 2.1.287.
What to do
Measure every tool description and your server instructions today. Keep warnings and limits inside the first 2,048 characters, because users on the stable channel still run a release from before this change.
Write the description for the smallest window your users run. The bigger window is a bonus, never the plan.
On Wednesday 7 October 2026 at 23:57 UTC, @modelcontextprotocol/inspector 2.10.0 shipped with two bins, mcpdo and mcp-inspector, and npx @modelcontextprotocol/inspector stopped working. Version 2.10.1 followed at 05:22 UTC on Thursday 8 October with an inspector bin alias, and npm now marks 2.10.0 deprecated.
What to do
If users start your MCP server with npx, keep one bin named exactly like the unscoped package name before you add a second one. Then test the plain npx your-package command in CI, not only the bin by name.
Your install line is part of your API. I would test it like one, on every release.
On Thursday 8 October 2026, Google Cloud introduced the Gemini agent at Gemini at Work 2026. Google says it can connect securely with any MCP server inside or outside a company network, and offers an enterprise tools registry where teams publish tools for the rest of the company.
What to do
If you run a remote MCP server for business users, check that it works with standard OAuth and write one plain paragraph per tool on what it reads and what it changes.
Every big platform wants its own tool shelf now. I'd make my server the easiest one on the shelf to approve.
On Friday 2 October 2026, the official TypeScript SDK shipped @modelcontextprotocol/server 2.3.0 as a minor release, then 2.3.1 on Monday 5 October. A server object now serves one connection at a time. Reuse one McpServer across stateless HTTP requests and the second request can fail with ALREADY_CONNECTED.
What to do
Search your code for an McpServer created outside the request handler. Move it inside, upgrade to 2.3.1, and send two requests at once before you ship.
Minor version. Major surprise. I'd test request number two before I trust request number one.
On Wednesday 7 October 2026, GitHub made local sandboxing generally available in Copilot CLI, the Copilot app and VS Code sessions using Agent Host. Tools and commands Copilot starts run with restricted access to files, network and credentials. GitHub says this covers local MCP servers where supported, and enterprises can enforce policies developers cannot weaken.
What to do
If you ship a local MCP server, list every path, host and credential it needs in your README, then run it under a Copilot local sandbox before your users do.
A local server used to inherit the whole laptop. Now it gets what the policy allows. Write down what you need, or expect to get nothing.
On Tuesday, 6 October 2026, SEP-2127 was merged into the MCP repository with status Final, on the Extensions Track. A remote server can now describe itself before anyone connects (name, version, transports and protocol versions) at <streamable-http-url>/server-card, with an optional AI Catalog at /.well-known/ai-catalog.json.
What to do
If you run a remote MCP server, publish a card at /server-card next to your endpoint and bump its version with every release.
The card tells an agent who you are. It can not tell it whether you answered a minute ago. That part still has to be measured.
On Monday, 5 October 2026, the MCP project merged a security guide for servers that run on your own machine. It names five threats (malicious servers, overcollecting servers, compromised dependencies, poisoned tool catalogs and vulnerable components) and calls running third-party servers in containers the most effective control.
What to do
Pin every local server to an exact version, give each one its own narrow token, and run third-party ones in a container that mounts only the project folder.
Every npx line in your config is a stranger you let in with your keys. Treat it like one.
On 6 October 2026 the MCP TypeScript SDK maintainers published CVE-2026-104850, rated High at CVSS 7.5. The OAuth client could send credentials to an authorization server chosen by the MCP server. Affected: @modelcontextprotocol/sdk 1.12.0 to 1.30.1 and @modelcontextprotocol/client 2.0.0 to 2.1.0; fixed in 1.31.0 and 2.2.0.
What to do
Upgrade to sdk 1.31.0 or client 2.2.0, pass expectedIssuer to the bundled providers, and rotate the client secret and revoke tokens for any client that may have touched an untrusted MCP server.
An OAuth client that lets the server choose the login page is a client that hands out keys on request.
On Monday, 5 October, SEP-3415 was opened as a draft Events extension for MCP. Servers would list event types with events/list and deliver them by polling, streaming, or webhooks signed with Standard Webhooks. The draft lists eight open questions and has no reference SDK yet.
What to do
If your server already emits webhooks, read the draft and comment on the open questions now; hold off on shipping until a reference SDK lands.
Events turn a tool list into an agent that reacts. I'd watch the replay rules closely, because that is where a missed event goes unnoticed.
On Wednesday, 30 September, Cloudflare opened a closed beta of Monetization Gateway. Sellers can charge agents for sites, APIs, MCP tools or datasets through HTTP 402 and x402, settled in USDC on Base via Coinbase's x402 facilitator. For now the beta is limited to eligible U.S. participants on both sides of the payment.
What to do
If you sell API or MCP access to U.S. customers and already proxy through Cloudflare, apply from the dashboard and price one tool per call before you price the whole server.
Per-call pricing changes tool design. A tool that needs five calls to finish one job now costs the agent five times as much, so merge your chatty tools.
CVE records published on Tuesday, 29 September 2026 cover MetaMCP up to and including 2.4.22. CVE-2026-79538 (CVSS 9.8 per CISA) allows code execution through /mcp-proxy/server/stdio. CVE-2026-79537 (9.1) lets a caller use another tenant's session; the record says /metamcp/health/sessions hands out the session ids without authentication.
What to do
If you run MetaMCP, block /mcp-proxy/ at your reverse proxy, turn off open sign-up, keep it off the public internet and rotate the secrets in its environment.
A gateway is the one box that knows every key. Patch it like one, and when there is no patch, hide it like one.
SEP-3392, filed 25 September 2026, sorts MCP extensions into experimental-ext-, beta-ext- and ext- names. Apps and auth would start as stable, skills and tasks as beta. A second proposal, SEP-3398, proposes releases every March and September, starting in March 2027.
What to do
List the extensions your server and client advertise today, and note which would land as experimental, so a rename does not break clients overnight.
Labels are cheap until a client trusts them. Then a renamed extension is an outage.
SEP-2809 is a draft that still has no sponsor; it had new review activity on 30 September 2026. A server would publish an Ed25519-signed document at /.well-known/mcp-attestation, and the host would check it against a trust root it pinned before allowing tools. Every admission and denial would go into a hash-chained, append-only audit log.
What to do
Nothing to do yet; watch for SEP-2809 to find a sponsor.
Every protocol ends up rebuilding the passport office. The only open question is who holds the stamp.
An independent count published on 2 October 2026 sent an initialize request to 186 remote MCP endpoints from the official registry on 1 and 2 October. Of the 72 that returned a protocol version, 7 (9.7%) negotiated the newest revision, 2026-07-28, and the rest answered with one of four older ones; 86 sat behind a sign-in.
What to do
Send your own server an initialize request right now and see which revision it answers with. Keep the older revisions working until your clients have actually moved.
The newest spec gets the blog posts. The older ones still do most of the talking.
In SEP-2127 (MCP Server Cards, metadata a client reads before it connects), a 24 September comment said the card is not at a generic well-known address and needs an AI Catalog to point to it. SEP-1913 (trust and sensitivity annotations) is being split into small extensions in an experimental repository, with an optional evidenceRef field that carries a digest, a canonicalization and a schema; both proposals are still open.
What to do
Nothing to do yet; watch for where SEP-2127 settles the card before you publish one at a fixed address.
The rules for who gets found are being written right now in a GitHub thread. I'd rather read it this week than learn the result later.
Microsoft introduced Work IQ Developer Tools (WIQD) on 30 September 2026: a plugin in the Agent Plugins format, which packages skills and MCP server configurations, can be imported, validated and packaged for Microsoft Copilot. From there it can be shared with named users or a whole tenant, or sent through Partner Center for certification.
What to do
If your plugin uses the Agent Plugins format, run it through the WIQD import and validate steps this week and fix what they flag. Then call your MCP server from Copilot yourself, because a package that validates is not the same as a server that answers.
A new client is a new chance to be found, and a new way to fail in public. I'd test before I'd announce.
At DevDay on 29 September 2026, OpenAI said ChatGPT can suggest a relevant plugin during a conversation and let people connect it on the spot. Its plugin guidelines say plugins with strong real-world utility and high user satisfaction may get enhanced distribution, such as directory placement or proactive suggestions.
What to do
Measure how fast your MCP server answers and how often its calls fail, and fix the worst tool this week. Satisfaction is one of the stated signals, and nobody is satisfied by a timeout.
Google gave you a page two. ChatGPT doesn't. You're either recommended, or you don't exist.
On 25 September 2026 Anthropic opened its directory submission portal to developers on paid Claude plans. You can submit a single MCP connector or a plugin bundle of MCP servers and skills hosted on GitHub; each submission is checked and safety-scanned as soon as you submit, and after launch you see installs by product surface and version.
What to do
If people already use your MCP server, call each tool once from a clean client today and fix anything slow or broken. Then submit it as a connector or a plugin bundle and let the scan tell you what else to fix.
A directory listing gets you the first click. Only a server that answers gets you the second.