Swagger MCP Bridge
What it does
Spring Boot example that exposes SpringDoc OpenAPI operations as safe MCP tools.
- Connects 在你的电脑上运行
- Account or key No key needed
- Category 开发者工具
安装
选择你使用的 AI 应用,然后复制设置。
通过 3/3 项检查 · 在你的电脑上运行
第一步:你用哪个 AI?
选择你用的应用,步骤会显示在下方。
可连接 Claude,但需要先准备
需要先在电脑上安装 Docker(只需一次) · 仅限电脑上的 Claude 应用(Windows 或 Mac),网页和手机不可用
打开电脑版 Claude 应用的设置文件
SettingsDeveloperEdit Config
粘贴这些设置并保存文件
{ "mcpServers": { "swagger-bridge": { "command": "docker", "args": [ "run", "-i", "--rm", "ghcr.io/neo1228/swagger-mcp-bridge-example:0.1.0" ] } } }退出 Claude 再重新打开
试一试
用 Swagger MCP Bridge 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Claude 说正在使用 Swagger MCP Bridge,或回答中出现 Swagger MCP Bridge 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同
ChatGPT 暂时无法连接 Swagger MCP Bridge
Swagger MCP Bridge 需要安装在你的电脑上,而 ChatGPT 只能连接在网上运行的服务器
你可以改用这些方式
A · 使用能连接 Swagger MCP Bridge 的应用
- 需要先在电脑上安装 Docker(只需一次) · 仅限电脑上的 Claude 应用(Windows 或 Mac),网页和手机不可用
- 需要先在电脑上安装 Docker(只需一次)
- 需要先在电脑上安装 Docker(只需一次)
- 需要先在电脑上安装 Docker(只需一次)
B · 开发者工具 类中能连接 ChatGPT 的其他服务器
从同类中使用较多的里挑选,仅限用地址即可连接、无需密钥的
- Sentry✓ ChatGPT
- Compiler Explorer✓ ChatGPT
- GitDiagram✓ ChatGPT
可连接 Claude Code,但需要先准备
需要先在电脑上安装 Docker(只需一次)
打开终端
Mac:按 Command+空格,输入 Terminal · Windows:打开 PowerShell
粘贴这条命令并按 Enter
claude mcp add --scope user swagger-bridge -- docker run -i --rm ghcr.io/neo1228/swagger-mcp-bridge-example:0.1.0在 Claude Code 中输入 /mcp
应能看到 Swagger MCP Bridge 在列表中并显示已连接
试一试
用 Swagger MCP Bridge 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Claude Code 说正在使用 Swagger MCP Bridge,或回答中出现 Swagger MCP Bridge 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同
可连接 Cursor,但需要先准备
需要先在电脑上安装 Docker(只需一次)
在 Cursor 中以 Agent 模式打开 AI 聊天
试一试
用 Swagger MCP Bridge 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Cursor 说正在使用 Swagger MCP Bridge,或回答中出现 Swagger MCP Bridge 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Cursor 帮助页面核实 · 你的界面可能因版本略有不同
可连接 VS Code,但需要先准备
需要先在电脑上安装 Docker(只需一次)
在 VS Code 中以 Agent 模式打开 AI 聊天
试一试
用 Swagger MCP Bridge 告诉我你能帮我做什么,然后做一件给我看✓ 如果 VS Code 说正在使用 Swagger MCP Bridge,或回答中出现 Swagger MCP Bridge 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 VS Code 帮助页面核实 · 你的界面可能因版本略有不同
开发者专用:终端命令和 JSON 设置
安装在电脑上
Claude Code
claude mcp add --scope user swagger-bridge -- docker run -i --rm ghcr.io/neo1228/swagger-mcp-bridge-example:0.1.0Claude Desktop (claude_desktop_config.json)
{
"mcpServers": {
"swagger-bridge": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/neo1228/swagger-mcp-bridge-example:0.1.0"
]
}
}
}Cursor (~/.cursor/mcp.json)
{
"mcpServers": {
"swagger-bridge": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/neo1228/swagger-mcp-bridge-example:0.1.0"
]
}
}
}VS Code (mcp.json)
{
"servers": {
"swagger-bridge": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/neo1228/swagger-mcp-bridge-example:0.1.0"
]
}
}
}这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
一览
哪些应用可以使用它
依据这个服务器的连接方式和各应用自己的文档,检查于 2026年11月10日。
在你的电脑上运行,适用于
ClaudeGooseLM StudioMsty StudioCherry StudioCursorVS Code + GitHub CopilotVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantClaude CodeCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatRaycast
它能做什么
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
基础检查i
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月13日
- ✓ 通过
作者与仓库一致
注册名称已通过 GitHub 账户 Neo1228 验证,该账户拥有此仓库
- ✓ 通过
仍在维护
代码最后更新于3个月前
- ✓ 通过
有许可证
使用 Apache-2.0
- – 不适用
端点有响应
它在你的电脑上运行,没有互联网端点
- – 不适用
使用 OAuth 登录
在你的电脑上运行,改用本地设置
- … 尚未检查
工具描述中没有隐藏指令
它在你的电脑上运行;我们只能读取互联网端点的工具列表
讨论
提问、分享技巧或报告无法使用。按“有帮助”数量排序。
用户评价
还没有评价
与“开发者工具”中的其他服务器比较i
- 星标数多于此分类中其他 5819 个服务器的 83%
- 发布到注册中心的时间比此分类中其他 5819 个服务器的 10% 更新
- 基础检查结果等于或优于此分类中其他 5819 个服务器的 100%
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
常见问题
Swagger MCP Bridge 安全吗?
它在 3 项适用的基础检查中通过了 3 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
它需要哪些密钥或代码?
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在我的电脑上运行还是在云端?
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
它还在维护吗?
GitHub 仓库最后更新于3个月前,最新版本(0.1.0)于6个月前进入注册中心。
哪些应用可以使用它?
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Docker。
同类别的其他选择
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:↓ 每周下载量i★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
GitHub
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
Serena MCP: the IDE for your agent
A powerful toolkit for coding, providing semantic retrieval and editing capabilities.
作者的 README
打开 README
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Swagger MCP Bridge
Turn any SpringDoc-powered Spring Boot API into a production-ready MCP gateway.
Swagger MCP Bridge discovers your OpenAPI operations, publishes them as safe MCP tools, and adds a smart gateway layer for API discovery, validation, response shaping, and multi-step workflow orchestration.
Naming and Coordinates
This repository intentionally separates the public surfaces so each one reads naturally in its own ecosystem:
| Surface | Name |
|---|---|
| Project / docs | Swagger MCP Bridge |
| Maven dependency | io.github.neo1228:openapi-mcp-spring-boot-starter |
| Official MCP Registry server | io.github.Neo1228/swagger-mcp-bridge |
| Runnable example image | ghcr.io/neo1228/swagger-mcp-bridge-example: |
| Spring configuration prefix | swagger.mcp.* |
The project name keeps the established bridge branding, while the Maven artifact uses the neutral OpenAPI/Spring Boot starter coordinate expected by Java consumers. The registry server and GHCR image identify the packaged runnable example used by MCP directories.
Why Swagger MCP Bridge
Most MCP API bridges stop at a thin tool wrapper. Swagger MCP Bridge is designed as a runtime gateway: it exposes your existing Spring controllers to LLM clients while preserving contracts, guardrails, and operational visibility.
What You Get
- Zero-boilerplate discovery of SpringDoc OpenAPI operations from your running Spring app
- Automatic MCP tool registration for discovered API operations
- Smart-context gateway tools:
meta_get_api_capabilities,meta_validate_api_call,meta_discover_api_tools,meta_describe_api_tool,meta_list_api_groups,meta_plan_api_workflow,meta_invoke_api_workflow,meta_invoke_api_by_intent - API catalog and workflow layer for capability inspection, preflight validation, grouped exploration, dry-run planning, and sequential execution
- Rich MCP input schemas generated from OpenAPI constraints: required fields, enums, numeric/string/object limits, examples, and deprecation hints
- Response shaping with JSONPath projection and summarization controls
- Execution guardrails: required argument validation, unresolved path-template protection, and safe
_headersfiltering - Structured MCP error responses with stable codes such as
INVALID_ARGUMENT,SECURITY_DENIED,WORKFLOW_ERROR, andHTTP_DISPATCH_FAILED - Java 17 bytecode with CI coverage on Java 17, 21, and 25
- Optional virtual-thread HTTP dispatch on Java 21+ runtimes, with automatic platform-thread fallback on Java 17
- Production guardrails for dangerous operations:
_confirm, blocked paths, role checks, audit logs, and structured client errors
Architecture
graph TD
User([User / LLM Client]) <--> MCP[MCP Client / Claude Desktop] MCP <--> Bridge[Swagger MCP Bridge /starter/] Bridge --> Catalog[Operation Catalog /groups + contracts/] Bridge --> Workflow[Workflow Orchestrator /plan + dry-run + execute/] Bridge <--> Docs[SpringDoc OpenAPI /v3/api-docs] Bridge <--> API[Your Spring Controller /hello]
Quick Start
1. Create a Spring Boot app
Use:
- Java 17+
- Spring Boot 3.5.x
- Spring Web
2. Add dependencies
Gradle (build.gradle.kts):
plugins {
id("org.springframework.boot") version "3.5.14" id("io.spring.dependency-management") version "1.1.7" java }
java { sourceCompatibility = JavaVersion.VERSION_17 }
repositories { mavenCentral() }
dependencies { implementation("org.springframework.boot:spring-boot-starter-web") implementation("org.springdoc:springdoc-openapi-starter-webmvc-api:2.8.17") implementation("io.github.neo1228:openapi-mcp-spring-boot-starter: ") }
The Maven artifact intentionally uses the neutral OpenAPI name rather than Swagger branding:
io.github.neo1228:openapi-mcp-spring-boot-starter
Maven (pom.xml):
<properties>
0.1.0-SNAPSHOT
org.springframework.boot spring-boot-starter-web
org.springdoc springdoc-openapi-starter-webmvc-api 2.8.17
io.github.neo1228 openapi-mcp-spring-boot-starter ${openapi-mcp.version}
Use a release version (for example 0.1.0) when consuming from a remote artifact repository.
3. Add one controller
import io.swagger.v3.oas.annotations.Operation;
import org.springframework.web.bind.annotation.GetMapping; import org.springframework.web.bind.annotation.RequestParam; import org.springframework.web.bind.annotation.RestController;
import java.util.Map;
@RestController public class HelloController {
@Operation(operationId = "getHello", summary = "Get greeting message") @GetMapping("/hello") public Map hello(@RequestParam(defaultValue = "world") String name) { return Map.of("message", "Hello " + name); } }
4. Add configuration (application.yml)
spring:
ai: mcp: server: protocol: STREAMABLE_HTTP streamable-http: mcp-endpoint: /mcp
swagger: mcp: enabled: true api-docs-path: /v3/api-docs tool-name-prefix: api_
5. Run and verify
- Start app:
./gradlew bootRunor./mvnw spring-boot:run - Verify OpenAPI:
http://localhost:8080/v3/api-docs - Verify MCP endpoint:
http://localhost:8080/mcp - Connect from an MCP client
Generated tool names follow (example: api_gethello).
MCP Client Workflow
This starter exposes direct API tools and a meta-tool layer so general MCP clients can work with large APIs without guessing tool names upfront:
meta_get_api_capabilitiesreturns API catalog stats, available gateway tools, orchestration features, safety policy, and response controls.meta_list_api_groupssummarizes the exposed API catalog by OpenAPI tag/group.meta_discover_api_toolsfinds relevant operations for a natural-language request.meta_describe_api_toolreturns the selected tool's method/path, parameters, required arguments, request body schema, risk flags, and full MCP input schema.meta_validate_api_callvalidates one generated API tool call without dispatching HTTP, including required arguments, risky-operation confirmation, and dispatch preview.meta_plan_api_workflowturns a workflow goal into a deterministic candidate step plan with contracts and risk flags.meta_invoke_api_workflowdry-runs or executes multiple generated API tools sequentially.meta_invoke_api_by_intentcan select and invoke the best matching operation when the client already has enough arguments.
The configured tool-name-prefix is still applied, so the default generated names are api_meta_get_api_capabilities, api_meta_validate_api_call, api_meta_list_api_groups, api_meta_discover_api_tools, api_meta_describe_api_tool, api_meta_plan_api_workflow, api_meta_invoke_api_workflow, and api_meta_invoke_api_by_intent.
When a tool call is rejected, the text content remains human-readable and structuredContent.error gives clients a stable machine contract:
{
"error": { "code": "INVALID_ARGUMENT", "message": "Missing required argument(s): path parameter: orderId", "status": 400, "retryable": false, "details": { "toolName": "api_getorder" } } }
Recommended client loop:
- Call
api_meta_get_api_capabilitiesonce to learn the gateway features and safety policy. - Use
api_meta_discover_api_toolsorapi_meta_list_api_groupsto find candidate operations. - Use
api_meta_describe_api_toolfor exact argument schema. - Use
api_meta_validate_api_callbefore risky or generated calls. - For multi-step work, call
api_meta_plan_api_workflow, thenapi_meta_invoke_api_workflowwithdryRun=true, then execute withdryRun=falseonly after validation is clean.
Workflow execution is intentionally safe by default:
meta_validate_api_callandmeta_invoke_api_workflowdry-runs validate tool names, arguments, required fields, dispatch paths, and risk flags before dispatching HTTP.- A workflow step has
{ "id": "...", "toolName": "...", "arguments": { ... } }. - Later steps can read previous structured results with JSONPath interpolation:
${create:$.order.id}. - If the whole argument value is a template, the resolved raw value is passed through. If a template is embedded in a longer string, the value is stringified.
- Recursive meta-tool orchestration is blocked; workflow steps can invoke generated API operation tools only.
- Risky HTTP methods still require the configured
_confirmtoken even inside a workflow.
Example validation payload:
{
"toolName": "api_getorder", "arguments": { "orderId": "order-1" } }
Example workflow payload:
{
"dryRun": false, "steps": [ { "id": "create", "toolName": "api_createorder", "arguments": { "body": { "id": "order-1", "item": "shoe" }, "_confirm": "CONFIRM" } }, { "id": "read", "toolName": "api_getorder", "arguments": { "orderId": "${create:$.order.id}" } } ] }
For larger APIs, set swagger.mcp.smart-context.gateway-only=true to expose only this gateway/meta layer instead of registering every operation as a top-level MCP tool.
Local Development Install
If the artifact is not published to a remote registry yet:
- Build and publish to local Maven cache:
./gradlew publishToMavenLocal
- In your consumer app:
- add
mavenLocal()repository - use version
0.1.0-SNAPSHOT(or your chosen local version)
Key Configuration
swagger.mcp.enabled: enable/disable bridge (defaulttrue)swagger.mcp.api-docs-path: OpenAPI docs path (default/v3/api-docs)swagger.mcp.tool-name-prefix: tool name prefix (defaultapi_)swagger.mcp.smart-context.gateway-only: expose only meta toolsswagger.mcp.execution.virtual-threads-enabled: run outbound API dispatch through virtual threads when the current runtime supports them (defaulttrue; safely falls back on Java 17)swagger.mcp.execution.allowed-argument-headers: optional allowlist for dynamic_headerspassed by MCP clientsswagger.mcp.execution.blocked-argument-headers: denylist for dynamic_headers; defaults block hop-by-hop/transport-sensitive headers likeHost,Content-Length,Connection, andTransfer-Encodingswagger.mcp.security.require-confirmation-for-risky-operations: require_confirmtoken for risky methods
For risky HTTP methods (POST, PUT, PATCH, DELETE), default policy requires _confirm=CONFIRM. The adapter also validates missing required path/query/header/body arguments before dispatching HTTP, so MCP clients get a clear tool error instead of a malformed API call.
Compatibility Matrix
| Starter | Java | Spring Boot | springdoc-openapi | Spring AI BOM |
|---|
| 0.1.