Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Comprehensive guide for building Model Context Protocol (MCP) servers with support for tools, resources, prompts, and authentication. Use when: (1) Creating custom MCP servers, (2) Integrating external APIs with Claude, (3) Building tool servers for specialized domains, (4) Creating resource providers for documentation, (5) Implementing authentication and security
.claude/skills/aiskillstore-mcp-builder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -27% | 0% |
Build production-ready MCP servers with tools, resources, prompts, and interactive widgets using mcp-use.
Decompose user requests into tools, widgets, and resources. Decide what needs UI vs text.
Read design-and-architecture.md: when planning what to build, deciding tool vs widget, or designing UX flows.
server.tool(), server.resource(), server.prompt() coderesources/ foldertypescriptimport { MCPServer, text, object, markdown, html, image, widget, error, } from "mcp-use/server"; import { z } from "zod"; const server = new MCPServer({ name: "my-server", version: "1.0.0" }); // Tool server.tool( { name: "my-tool", description: "...", schema: z.object({ param: z.string().describe("...") }), }, async ({ param }) => text("result"), ); // Resource server.resource( { uri: "config://settings", name: "Settings", mimeType: "application/json" }, async () => object({ key: "value" }), ); // Prompt server.prompt( { name: "my-prompt", description: "...", schema: z.object({ topic: z.string() }), }, async ({ topic }) => text(`Write about ${topic}`), ); server.listen();
Response helpers: text(), object(), markdown(), html(), image(), audio(), binary(), error(), mix(), widget()
Server methods: server.tool(), server.resource(), server.resourceTemplate(), server.prompt(), server.listen()
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 29,832 | 23,134 | -22% | 1 | 1 | 0% | 4,263 | 4,215 | -1% | 0 | 0 | — |
case-02 | fail→pass | 28,592 | 22,672 | -21% | 1 | 1 | 0% | 5,197 | 4,309 | -17% | 0 | 0 | — |
case-05 | pass→pass | 17,973 | 15,838 | -12% | 1 | 1 | 0% | 2,292 | 2,751 | +20% | 0 | 0 | — |
case-03 | fail→pass | 33,557 | 26,442 | -21% | 1 | 1 | 0% | 6,119 | 6,349 | +4% | 0 | 0 | — |
case-04 | pass→pass | 9,809 | 12,738 | +30% | 1 | 1 | 0% | 1,845 | 2,252 | +22% | 0 | 0 | — |
case-06 | pass→pass | 10,180 | 12,089 | +19% | 1 | 1 | 0% | 1,909 | 2,970 | +56% | 0 | 0 | — |
case-07 | fail→pass | 25,850 | 7,842 | -70% | 1 | 1 | 0% | 2,762 | 2,038 | -26% | 0 | 0 | — |
case-08 | fail→pass | 13,129 | 2,868 | -78% | 1 | 1 | 0% | 1,607 | 1,170 | -27% | 0 | 0 | — |
case-09 | fail→pass | 18,917 | 7,474 | -60% | 1 | 1 | 0% | 2,778 | 1,911 | -31% | 0 | 0 | — |
case-10 | fail→pass | 21,191 | 14,369 | -32% | 1 | 1 | 0% | 3,062 | 3,539 | +16% | 0 | 0 | — |
case-11 | fail→pass | 25,987 | 7,049 | -73% | 1 | 1 | 0% | 2,761 | 2,030 | -26% | 0 | 0 | — |
case-12 | fail→pass | 20,994 | 16,210 | -23% | 1 | 1 | 0% | 2,446 | 2,318 | -5% | 0 | 0 | — |
case-13 | fail→pass | 15,460 | 10,992 | -29% | 1 | 1 | 0% | 1,936 | 1,685 | -13% | 0 | 0 | — |
case-14 | fail→pass | 14,730 | 5,987 | -59% | 1 | 1 | 0% | 2,136 | 1,673 | -22% | 0 | 0 | — |
case-15 | pass→pass | 8,324 | 2,190 | -74% | 1 | 1 | 0% | 1,598 | 946 | -41% | 0 | 0 | — |
case-16 | fail→pass | 16,894 | 8,204 | -51% | 1 | 1 | 0% | 2,141 | 1,148 | -46% | 0 | 0 | — |
case-17 | pass→pass | 8,432 | 10,332 | +23% | 1 | 1 | 0% | 1,687 | 1,523 | -10% | 0 | 0 | — |
case-18 | fail→pass | 18,204 | 4,395 | -76% | 1 | 1 | 0% | 2,241 | 1,343 | -40% | 0 | 0 | — |
case-19 | fail→pass | 13,907 | 6,809 | -51% | 1 | 1 | 0% | 1,531 | 910 | -41% | 0 | 0 | — |
case-20 | fail→pass | 13,236 | 2,577 | -81% | 1 | 1 | 0% | 1,511 | 1,055 | -30% | 0 | 0 | — |
case-21 | fail→pass | 16,249 | 1,339 | -92% | 1 | 1 | 0% | 1,811 | 782 | -57% | 0 | 0 | — |
case-22 | fail→pass | 14,879 | 6,940 | -53% | 1 | 1 | 0% | 1,571 | 825 | -47% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +77 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.