Youtube MCP Server 50759788

MCPAI代理

提供方 ctgunn

AI代理 stdio · 本地进程 Token 社区 / 第三方
查看接入方式

一个符合MCP协议的服务器,封装了YouTube Data API并提供可搜索的工具,用于OpenAI Agent Builder工作流。

一句话选型结论

按资料字段整理

Youtube MCP Server 50759788 属于AI代理类 MCP 服务器,stdio · 本地进程,Token,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类AI代理 传输方式stdio · 本地进程 鉴权Token 部署方式本地部署
GitHub 星标 0 仓库热度,无公开仓库时显示未标注
工具数 5 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

AI代理
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量5暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

适用场景

MCP协议Python本地部署YouTubeAPIOpenAI搜索工具网络服务API封装

安装与配置

stdio · 本地进程 Token
客户端在本机拉起子进程,数据不出本机
python3 -m venv .venv

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Python,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后客户端会列出约 5 个工具,随便调一个只读工具确认能返回结果。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
给代理接外部能力

让代理在任务中途调用搜索、数据库等真实工具。

多步任务编排

把一串操作拆成步骤,由代理按顺序执行。

自动化重复流程

把固定流程交给代理跑,减少手工操作。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 资料标注 5 个工具,职责相对集中,接入后不容易误操作。
  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 使用前要准备好Token对应的凭据,并按最小权限配置。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在AI代理场景里让模型直接调用外部能力的人。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 不方便管理密钥或第三方授权的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Youtube MCP Server 50759788当前 5 stdio · 本地进程 本地部署 有 2026-05-17
Filesystem 560c4363 13 未标注 本地部署 无 2026-05-17
Server Everything (everything) 11 stdio · 本地进程 本地部署 有 2026-05-17
Time 7365cb6d 2 SSE · 服务端推送 远程部署 有 2026-05-18
Everything Claude Code 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Openmemory 557e1585 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

AI代理
字段来自项目仓库与公开索引,中文说明由本站标注
提供方ctgunn上游标注的维护方 ctgunn
分类AI代理本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式Token远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · Python本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2026-05-17上游最后核验时间

详细介绍

12.8 KB
来自项目仓库与上游目录的原文,未做删改

youtube-mcp-server

An MCP-compliant server that wraps the YouTube Data API and exposes tools for
remote MCP clients, including OpenAI Agent Builder workflows. The current
hosted provider adapter runs on Google Cloud Run; the shared platform contract
keeps the overall model portable.

Start here

This is the repository's getting-started page. Choose the path that matches
what you want to do first.

Goal Start here Outcome
Explore, develop, or test locally Run locally A local MCP endpoint with no cloud account required
Test durable sessions locally Hosted-like local path The local server plus Redis in a container
Operate your own hosted instance Deploy your fork to GCP Cloud Run, durable session storage, and GitHub Actions releases
Understand the system before changing it Architecture guide The request lifecycle and module responsibilities

Run locally

Prerequisites

  • Python 3.11 or newer
  • pip
  • make
  • A YouTube Data API key only when calling live YouTube tools

Create an isolated environment and install the project:```bash
git clone https://github.com/<your-account>/youtube-mcp-server.git
cd youtube-mcp-server
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -e '.[dev]'
cp .env.example .env.local


Edit `.env.local` for your machine. It is private configuration and must never
be committed. Add `YOUTUBE_API_KEY` only when you need live API calls.

Run the full quality gate, then start the server:```bash
make quality
bash scripts/dev_local.sh

In another terminal, verify the process:```bash
curl http://127.0.0.1:8080/health
curl http://127.0.0.1:8080/ready


The minimal local runtime path does not require cloud provisioning, Docker, or
Redis. For MCP examples and local troubleshooting, read the
local development guide.

## Verify the default MCP tool catalog

Run the deterministic Layer 4 catalog check when changing default tool
registration, metadata, dispatch, or local fixtures:```bash
make test-tools

It discovers the catalog through the public MCP tools/list route and calls
every discovered tool through tools/call. The check uses no credential, makes
no outbound YouTube request, and performs no destructive external mutation.
It is a release-verification check for the default catalog, not proof that a
configured runtime can access YouTube.

Verify configured YouTube runtime capability

Run the deterministic configured-runtime matrix when changing YouTube runtime
configuration, credential selection, or the live execution boundary:```bash
make test-runtime


It discovers public YouTube families through MCP and checks API-key executable,
OAuth-required, and unavailable-capability outcomes through the same public
route. It uses controlled request construction, makes no outbound YouTube
request, and does not need a real credential. This verification is distinct
from both `make test-tools` catalog coverage and the optional live smoke below.

## Run the optional read-only YouTube live smoke

The live smoke is manual and opt-in. It consumes YouTube quota, depends on
upstream availability, and must be run only by an approved operator with a
real API key. It invokes all 12 explicitly reviewed API-key public-read endpoint
tools with bounded public fixtures. It excludes mutations, uploads, downloads,
deletes, ratings, reports, moderation actions, OAuth-required, owner-only,
retrieval, and baseline-server operations. The `comments_list` check derives one
public parent-comment identifier in memory from the approved
`commentThreads_list` fixture; it never prints that identifier or any response body.```bash
# In .env.local (or .env):
YOUTUBE_API_KEY=&#x27;operator-supplied-value&#x27;
RUN_YOUTUBE_LIVE_SMOKE=1

# Then, from the repository root:
make test-live-smoke

Without both prerequisites, the command fails before constructing the app or
making a live request. Its output redacts credentials, authorization material,
request inputs, headers, and upstream response bodies. It is not included in
make test, make quality, CI, or automated deployment gates. This local
configured-runtime smoke does not verify a running remote MCP endpoint; that
operator workflow is reserved for OPS-405.

The command sources .env.local first and falls back to .env when the local
file is absent. Both files are ignored by Git; never commit real credentials.

Run the complete repository test suite with:```bash
make test


The credential-gated live smoke remains separate from the complete repository
test suite and must not be included in `make test-tools`.

## Run the optional remote MCP read-only live smoke

This fourth verification workflow is manual and opt-in. It connects to an
already running remote MCP endpoint, initializes a remote session, discovers
the deployed catalog, and invokes only the 12 explicitly reviewed API-key
public-read tools through the public transport. It validates the remote
endpoint&#x27;s authorization, session, selected handlers, and live YouTube path;
it does not instantiate the local application as a substitute.

Use only an approved staging or production endpoint after deployment. The
workflow consumes YouTube quota and depends on both remote endpoint and
upstream availability. It is never included in `make test`, `make quality`,
CI, or automated deployment gates.```bash
# In .env.local (or .env):
RUN_REMOTE_MCP_LIVE_SMOKE=1
REMOTE_MCP_URL=&#x27;approved-remote-mcp-url&#x27;
# Set MCP_AUTH_TOKEN only if the selected remote endpoint requires bearer authorization.
MCP_AUTH_TOKEN=&#x27;operator-supplied-value&#x27;
REMOTE_MCP_AUTH_REQUIRED=1

# Then, run both reviewed remote suites from the repository root:
make test-remote-live-smoke

The command fails before any remote request when opt-in authorization or the
remote endpoint is absent, and it requires MCP_AUTH_TOKEN before a request
when REMOTE_MCP_AUTH_REQUIRED=1. It retains the remote MCP session only in
memory and prints a compact selected-tool report plus passed, excluded, and
request-bound summaries. Its output never includes tokens,
authorization headers, session identifiers, request arguments, raw URLs,
exception text, or upstream response bodies.

To exercise either reviewed layer independently with the same authorization
settings, run:```bash
make test-remote-layer2-live-smoke
make test-remote-layer3-live-smoke


The Layer 3 suite is bounded to 10 API-key-compatible public fixtures. It excludes
OAuth caption/transcript workflows and higher-cost search/enrichment fan-out.

Use the correct verification boundary for the question at hand:

- `make test-tools` is deterministic catalog verification with no external calls.
- `make test-runtime` is deterministic configured-runtime verification with controlled requests.
- `make test-live-smoke` is local live smoke against a configured local application.
- `make test-remote-live-smoke` runs the independently bounded Layer 2 and Layer 3 remote smoke suites against a running MCP endpoint.
- `make test-remote-layer2-live-smoke` runs the reviewed Layer 2 API-key public-read suite only.
- `make test-remote-layer3-live-smoke` runs the reviewed Layer 3 API-key public-read suite only.

If remote live smoke reports a safe availability or actionable failure, record
only its safe tool/category summary. Review endpoint authorization, deployment
health, quota, and upstream status privately, then retry only after approval.

## Test hosted-like sessions locally

Use this optional path only when you need Redis-backed durable-session testing,
reconnect behavior, event replay, or hosted readiness behavior:```bash
./scripts/local_compose.sh up -d
LOCAL_SESSION_MODE=hosted bash scripts/dev_local.sh

Stop Redis when finished:bash ./scripts/local_compose.sh down

This remains local verification—Cloud Run, GCP IAM, and public reachability
are not involved. See the local infrastructure README
for the full hosted-like local verification path.

Deploy your fork to GCP

Forking copies the workflow definition, but a first deployment needs one-time
configuration in your GitHub repository and your GCP project. This is
intentional: project identity, Terraform state, and secret values do not belong
in this source repository.

1. Fork and enable Actions

  1. Fork this repository to an account or organization you control.
  2. Open the fork's Actions tab and enable Actions if GitHub prompts you.
  3. Use the fork's default branch—normally main—as the release branch.

hosted-deploy is manually dispatched by design. A hosted release is an
explicit, reviewed operator action rather than an automatic effect of a push.

2. Create the GCP bootstrap resources

In your GCP project, create the following once:

  1. A private, versioned GCS bucket for Terraform state.
  2. A regional Artifact Registry Docker repository.
  3. A GitHub OIDC Workload Identity Pool/provider restricted to your fork, plus
    a deployer service account that it may impersonate.
  4. Least-privilege access for that deployer to publish images, read/write
    Terraform state, reconcile the infrastructure, deploy Cloud Run, and manage
    runtime Secret Manager access bindings. OPS-402 additionally requires
    roles/logging.configWriter and roles/monitoring.alertPolicyEditor to
    create application metrics and alert policies.
  5. Secret Manager secret values for YOUTUBE_API_KEY and MCP_AUTH_TOKEN.

The release workflow manages the Cloud Run foundation, durable Redis session
path, and managed network bootstrap after these prerequisites exist. It does
not create your state bucket, image repository, identity federation, or secret
values.

3. Configure staging.tfvars

Replace the checked-in values in infrastructure/gcp/staging.tfvars with your
own project ID, region, service name, allowed browser origin, and managed
network names/CIDRs. The file may be committed only when it contains
non-sensitive resource configuration—never API keys, tokens, passwords, or
secret values.

4. Configure GitHub Actions

In the fork, open Settings → Secrets and variables → Actions.

Add repository variables:

Variable Value
GCP_PROJECT_ID Your GCP project ID
GCP_REGION Your Artifact Registry and Cloud Run region
GCP_SERVICE_NAME The service name from staging.tfvars
GCP_ARTIFACT_REGISTRY_REPOSITORY Your Artifact Registry Docker repository
GCP_TERRAFORM_VAR_FILE Normally staging.tfvars
GCP_TERRAFORM_STATE_BUCKET Your private, versioned Terraform state bucket

Add repository secrets:

Secret Value
GCP_WORKLOAD_IDENTITY_PROVIDER The full Workload Identity Provider resource name
GCP_DEPLOYER_SERVICE_ACCOUNT The deployer service-account email address

Keep YOUTUBE_API_KEY and MCP_AUTH_TOKEN in GCP Secret Manager. The
supported workflow uses short-lived identity federation, so do not create or
store a downloadable service-account key in GitHub or the repository.

5. Run the first release

In GitHub, open Actions → hosted-deploy → Run workflow. Select
main as both the workflow source and target_ref, then choose staging as
the target environment unless you are deliberately releasing another reviewed
revision.

The workflow runs safe preflight and make quality, publishes an immutable
image, reconciles Terraform, deploys Cloud Run, and runs hosted verification.
After it succeeds, download hosted-deploy- and keep the source
revision, image digest, Terraform outputs, deployment record, and verification
result as release evidence.

For detailed Terraform inputs, GCP roles, network/Redis behavior, and
break-glass recovery, continue with the
hosted deployment guide and the
GCP infrastructure README.

Documentation map

Document Use it for
docs/local-development.md Local runtime, MCP verification, test commands, and hosted-like local sessions
docs/architecture.md System design, request lifecycle, transport, tools, sessions, and runtime behavior
docs/engineering.md Quality rules, pull-request checks, contributor standards, and runtime-security configuration
docs/hosted-deployment.md GitHub Actions releases, bootstrap boundaries, evidence, failure handling, and break-glass recovery
infrastructure/local/README.md Redis-backed local dependency setup and troubleshooting
infrastructure/gcp/README.md Terraform inputs, GCP resources, Secret Manager, Redis, networking, and provider-specific operations
requirements/PRD.md Product goals, boundaries, and roadmap context
requirements/spec-kit-seed.md Feature-slice catalog and SpecKit branch numbering
requirements/tool-specs.md YouTube tool inventory and detailed tool-level requirements
docs/archive/README.md Historical material only; not operational guidance

The root README should stay focused on first use and finding the right guide.
Put detailed operational or technical reference material in the scoped document
that owns it.

常见问题

7 问
答案由本站依据条目字段整理,不引用上游原文
Youtube MCP Server 50759788 怎么安装?

在客户端按下面的安装命令接入:python3 -m venv .venv。装完在客户端配置里引用同一个包或端点。

Youtube MCP Server 50759788 怎么在客户端里配置?

它的传输方式未标注,在客户端 MCP 配置里按官方文档填写启动命令和参数,保存后重启客户端生效。

Youtube MCP Server 50759788 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Youtube MCP Server 50759788 需要 API Key 吗?

资料标注需要Token认证,使用前准备好对应凭据,并按最小权限配置。

Youtube MCP Server 50759788 能提供哪些能力?

资料显示约 5 个工具;能力标签包括 搜索工具、Python、API封装、OpenAI。具体工具清单以官方文档为准。

Youtube MCP Server 50759788 需要什么运行环境?

运行时为 Python;属于本地进程,需要在本机装好运行环境。

Youtube MCP Server 50759788 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-05-17。