--- title: "MCP vs. Skills: A False Debate" id: "3458" type: "insight" slug: "mcp-vs-skills" published_at: "2026-08-18T19:35:41+00:00" modified_at: "2026-08-18T19:35:42+00:00" url: "https://jetstream.security/insights/mcp-vs-skills/" markdown_url: "https://jetstream.security/insights/mcp-vs-skills.md" excerpt: "MCP gives an agent its tools and Skills give it the instructions. The teams winning with Claude run both, and governance is what keeps the pairing safe." taxonomy_insight_category: - "AI Advisory" taxonomy_insight_type: - "Editorial" taxonomy_insight_tags: - "JetStream Verified MCP" - "MCP Governance" - "MCP Servers" - "Shadow AI" --- AI Advisory # MCP vs. Skills: A False Debate MCP gives an agent its tools and Skills give it the instructions. The teams winning with Claude run both, and governance is what keeps the pairing safe. [JetStream Verified MCP](https://jetstream.security/insights/insight_tags/jetstream-verified-mcp/) [MCP Governance](https://jetstream.security/insights/insight_tags/mcp-governance/) [MCP Servers](https://jetstream.security/insights/insight_tags/mcp-servers/) [Shadow AI](https://jetstream.security/insights/insight_tags/shadow-ai/) [Tommy Hui](https://jetstream.security/author-bio/tommy-hui/) & Mike McGrail Editorial In whiteboarding sessions with customers across every size and vertical, the same question keeps coming up: now that Skills exist, is MCP still relevant? It is a fair question. Skills and the Model Context Protocol (MCP) both came from Anthropic, and they landed about a year apart. The newer one looks like it could stand in for the older. It can’t. The teams getting real work out of Claude run both together, and they are right to. The timeline explains the confusion. Anthropic announced MCP in November 2024, then released Skills in October 2025. In the year since, market chatter has cast the two as competing options. That reading is wrong. If MCP were on its way out, the rest of the industry would not be adopting it, and OpenAI and Google DeepMind both have (The New Stack, *Why the Model Context Protocol Won*). When Anthropic released MCP, it framed the protocol as a way to “connect AI assistants to the systems where data lives,” and to let developers “build secure, two-way connections between their data sources and AI-powered tools.” When Skills launched, Anthropic described them as “instructions, scripts, and resources” that an agent loads only when relevant, with the objective of making the agent “better at specialized tasks.” Here is the cleanest way to think about the difference. MCP servers provide the tools an agent can reach for. Skills provide the instructions for using those tools well. You can figure out how to operate a laser engraver on your own, but until someone hands you the user guide, you will not use it efficiently. ## Three workflows that show how Skills and MCP combine Three workflows make the relationship concrete. They are ordered by complexity, from a single Skill paired with a single MCP server to a stack that combines several of each. ### Use case 1: Research-to-deck An employee wants Claude to research a topic online and then produce a PowerPoint deck from the findings, using a predefined template. Workflow. Claude searches the web using the web search tool to gather recent data, reads the PowerPoint Skill to learn the slide-generation patterns and template format, and produces a .pptx file informed by the live research. Skills used: PowerPoint Skill. MCP servers used: Web search. The Skill is complementary to the MCP tool. They contribute to the same task in different ways: the MCP server brings the data in, the Skill formats the output. ### Use case 2: Dashboard-to-spreadsheet A team wants to export data from a web-based dashboard and land it in an Excel file with formulas and conditional formatting. Workflow. Claude uses the Chrome MCP tools to navigate to the dashboard, reads the page content, and extracts the tabular data through the accessibility tree or JavaScript execution. Claude then reads the XLSX Skill and builds a formatted spreadsheet with proper headers, formulas, and conditional formatting. Skills used: XLSX. MCP servers used: Chrome MCP. Two tools, two roles, one workflow. Pulling the data out is the MCP server’s job. Producing a spreadsheet that a finance partner can actually use is the Skill’s job. ### Use case 3: Third-party risk audit A risk management team wants to automate their third-party security audit workflow against a single vendor. Workflow. Claude navigates to the vendor’s security page via the Chrome MCP, extracts the published controls and certifications, searches the web for recent breach disclosures, pulls internal emails about the vendor via an email MCP, and reads an uploaded Software Bill of Materials (SBOM) or risk-assessment PDF using the PDF reading Skill. Claude then produces three outputs: an Excel risk matrix (XLSX Skill), a formal audit report (DOCX Skill), and an interactive risk dashboard artifact (front-end design Skill) that visualizes the findings as a React component. Skills used: PDF reading, XLSX, DOCX, front-end design. MCP servers used: Chrome MCP, email MCP, web search MCP. Three Skills and three MCP servers, stacked into a single audit run. This is the shape most enterprise workflows take once Skills and MCP servers are both in play. ## Stacking creates power. It also creates an operational surface. Stacking Skills with MCP servers is what makes the pattern useful. It is also what makes the operational surface bigger, not smaller. Every additional MCP server is another tool the agent can reach for. Every additional Skill is another set of instructions for how the agent uses those tools. The governance question follows directly: who can stack what, against which data, under whose identity, on whose behalf. In a previous post , we outlined the risks of [letting MCP sprawl continue](https://jetstream.security/insights/governing-mcp-sprawl/) without security or governance in place. To save you a tab, the top risks MCP servers create for enterprises include: - Expansion of shadow AI. - Vulnerabilities introduced by open-source and publicly available MCP servers. - Destructive tools made reachable by automated agentic workflows. - Ongoing privilege and permission management for agents and the workflows that orchestrate them. Skills do not change that picture. The same risks live in any environment that lets agents stack tools and instructions at runtime. ## What runtime governance actually does about it The JetStream SAIG Platform™ (Security-first AI Governance) is built for that environment. [JetStream AI Blueprints™](https://jetstream.security/jetstream-ai-blueprints/) create a versioned operational contract for how each agent is assembled: which MCP servers it can reach, which Skills it can run, and under whose identity it acts. The JetStream Identity Broker™ scopes that identity to the Blueprint and [revokes it when the contract changes](https://jetstream.security/insights/leaked-api-key-what-to-do/) . Discovery surfaces shadow MCP servers before they accumulate into shadow AI. Cost and permission controls run on the same plane. Two of the risks above have a more specific answer. [JetStream Verified MCP™](https://jetstream.security/insights/jetstream-verified-mcp-governance-layer-for-enterprise-ai-agents/) replaces open-source servers no one has inspected with hardened versions that are scanned, remediated, and cryptographically attested before an agent can call them. Tool-level filtering handles the destructive-tool problem directly: a team can enable a read tool and block a delete one, so an agent never holds the DELETE it never needed. Governance on this architecture does not slow agent adoption. It is what makes agent adoption at production scale possible. ## Stay ahead with our newsletter Stay informed on industry trends, expert analysis, and product updates. #### Explore more insights [See all Insights](/insights) [https://jetstream.security/insights/a-policy-is-not-a-control/](https://jetstream.security/insights/a-policy-is-not-a-control/) Aug 14, 2026 ###### A Policy Is Not a Control Approval happens in a moment, but the system keeps moving. Design control is the artifact that makes every AI change visible and every deviation provable, which… [https://jetstream.security/insights/a-policy-is-not-a-control/](https://jetstream.security/insights/a-policy-is-not-a-control/) [https://jetstream.security/insights/leaked-api-key-what-to-do/](https://jetstream.security/insights/leaked-api-key-what-to-do/) Aug 11, 2026 ###### Burt has a key problem. Make sure that you don't. One leaked API key shouldn't be able to stop your business. With a virtual key, it can't. [https://jetstream.security/insights/leaked-api-key-what-to-do/](https://jetstream.security/insights/leaked-api-key-what-to-do/) [https://jetstream.security/insights/ai-visibility-inventory/](https://jetstream.security/insights/ai-visibility-inventory/) Aug 7, 2026 ###### You Can’t Govern the AI You Can’t See Every AI standard starts with an inventory. Here's how to build the one your governance program depends on. [https://jetstream.security/insights/ai-visibility-inventory/](https://jetstream.security/insights/ai-visibility-inventory/) ## Top Articles 01 Press Release ### [JetStream Security Achieves FedRAMP Class D (High) Certification Through Second Front](https://jetstream.security/insights/jetstream-security-fedramp-high-certification/) July 22, 2026 02 Press Release ### [JetStream Releases ‘AI Kill Switch’ to Shut Down Individual Agents](https://jetstream.security/insights/jetstream-ai-kill-switch-agents/) July 27, 2026 03 Press Release ### [JetStream Announces Verified MCP Governance Layer for Enterprise AI Agents](https://jetstream.security/insights/jetstream-verified-mcp-governance-layer-for-enterprise-ai-agents/) July 13, 2026 ## Featured Experts [https://jetstream.security/author-bio/patrick-zeller/](https://jetstream.security/author-bio/patrick-zeller/) ### [Patrick E. Zeller](https://jetstream.security/author-bio/patrick-zeller/) General Counsel, Legal and Compliance Patrick has spent over twenty years advising Fortune 100 companies on privacy, cybersecurity, and data protection — including... [https://jetstream.security/author-bio/keith-weisman/](https://jetstream.security/author-bio/keith-weisman/) ### [Keith Weisman](https://jetstream.security/author-bio/keith-weisman/) Head of Forward Deployed Engineering Keith brings thirty years of hands-on cybersecurity and services leadership, beginning with enterprise security consulting at Accenture and... [https://jetstream.security/author-bio/tommy-hui/](https://jetstream.security/author-bio/tommy-hui/) ### [Tommy Hui](https://jetstream.security/author-bio/tommy-hui/) Head of Sales Engineering Tommy Hui has spent more than a decade leading sales engineering at security companies, including six years at...