能做什么
Hosting for apps you build with AI. Tested with Claude Code and Codex. Private beta: ask for access.
作者提供的原文
- 连接方式 在云端运行
- 账号或密钥 无需密钥
- 分类 开发者工具
可搭配使用
教 AI 做这类工作的 Skill,可与本项一起使用
andrej-karpathy-skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
graphify
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.
为什么排在这里
此项在“连接 AI 的应用”标签中,该标签不使用星标,按以下顺序排列
| 1. 在 Claude 连接器目录中 | 否 | |
|---|---|---|
| 2. 团队实测过 | 否 | |
| 3. 适合泰国用户 | 否 | |
| 4. 不计星标的使用量 | 15/100 | 热度 0 · 新鲜度 7 · 仍在响应 8 · 用户喜欢 +0 |
| 5. 登记日期 | 2026-10-19 |
用户喜欢: 0 · 最多加 5 分(每人 1 分),不能用钱买。
质量: 基础检查通过 4/4 项(仅展示,不用于排序)
安装
选择你使用的 AI 应用,然后复制设置。
用一句话让 AI 帮你安装
请帮我安装 fline。按照 https://mcp.popnx.com/i/sh-fline-fline.md 操作,先把命令给我看,等我同意后再运行。粘贴到能在你电脑上工作的 AI(如 Claude Code、Cursor)。它会读取我们的步骤,先给你看命令,等你同意。无需密钥,无需注册。 · 查看 AI 会读取的全文
通过 4/4 项检查 · 你的提问会经过开发者的服务器(mcp.fline.sh) · ✓ 命令已检查 2026年11月20日
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 · 约 2 分钟
不需要密钥,不用安装程序 · 在 claude.ai 或桌面应用设置一次,即可在聊天中使用 · 免费账户可添加 1 个 · Team/Enterprise 方案由所有者(Owner)添加
打开 Claude 的 Connectors 页面
CustomizeConnectors
点击“Add” → “Custom” → “Web”,填写这 2 项
把这里的内容复制到 Claude 中同名的输入框,然后点击 Continue
Name · 名称flineRemote MCP server URL · 地址https://mcp.fline.sh/mcp在 Claude 检测到的登录设置页点击 Continue · Authentication 按作者说明选择(Sign in now、Sign in when needed 或 No sign in),然后点击 Add
在聊天中开启
+Connectorsfline
试一试
用 fline 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 fline,或回答中出现 fline 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
能否连接 ChatGPT 取决于方案
完整 MCP:网页版 ChatGPT Business、Enterprise 或 Edu · Pro:以读取/获取权限连接,修改数据的工具可能用不了 · Plus:官方文档尚未确认,如果在插件页看到添加自定义 MCP 的选项可以试试 · 仅限网页
在网页版 ChatGPT 打开添加 MCP 的页面
Plugins+Add custom MCP server
填写名称和地址
输入名称,在 Connection 下粘贴地址,勾选风险提示,然后点击 Create as a plugin
Name · 名称flineConnection · 地址https://mcp.fline.sh/mcp在聊天中选择 fline
开启新聊天,输入前先从插件列表中选择 fline
试一试
用 fline 告诉我你能帮我做什么,然后做一件给我看✓ 如果 ChatGPT 说正在使用 fline,或回答中出现 fline 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 OpenAI 帮助页面核实 · 你的界面可能因版本略有不同
可连接 Claude Code · 约 2 分钟
不需要密钥,不用安装程序
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 --transport http --scope user fline https://mcp.fline.sh/mcp在 Claude Code 中输入 /mcp
应能看到 fline 在列表中并显示已连接
试一试
用 fline 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 fline,或回答中出现 fline 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 · 约 2 分钟
不需要密钥,不用安装程序
在 Cursor 中以 Agent 模式打开 AI 聊天
试一试
用 fline 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 fline,或回答中出现 fline 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 · 约 2 分钟
不需要密钥,不用安装程序
在 VS Code 中以 Agent 模式打开 AI 聊天
试一试
用 fline 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 fline,或回答中出现 fline 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
其他支持 MCP 的应用(如 Windsurf、Zed、Gemini CLI)添加方式各不相同,请在应用列表中查看
通过网页连接的应用使用的地址
https://mcp.fline.sh/mcpJSON 设置在下方“开发者专用”中
开发者专用:终端命令和 JSON 设置
通过网页连接
Claude Code
claude mcp add --transport http --scope user fline https://mcp.fline.sh/mcpCursor (~/.cursor/mcp.json)
{
"mcpServers": {
"fline": {
"url": "https://mcp.fline.sh/mcp"
}
}
}VS Code (mcp.json)
{
"servers": {
"fline": {
"type": "http",
"url": "https://mcp.fline.sh/mcp"
}
}
}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。
通过互联网连接,适用于
ClaudeChatGPTGooseLM StudioMsty StudioCherry StudioCursorVS Code + GitHub CopilotVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantClaude CodeCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatOpen WebUIRaycast
它能做什么
工具名称和描述与服务器在 2026年11月18日 报告的完全一致。这些内容由作者撰写,不是我们写的;我们以纯文本形式显示。
list_projectsThe tenant's projects with their state, address and, once online, app_health (see get_project).
create_projectCreate a project. Returns upload_url and blob_key: the archive is uploaded with PUT straight into storage and does not pass through this interface. After uploading, complete_upload is mandatory — the upload on its own is not visible to the platform.
delete_projectDelete the project. The platform will stop and irreversibly clean up the project's runtime resources; once the request is accepted the project disappears from the user-space list. Its saved passwords and keys are not deleted with it, and neither is a workspace database it used.
get_upload_urlIssue a new archive upload link for an existing project. Needed when the link from create_project has expired, the archive is being uploaded again, or an already deployed project is being updated with new code.
complete_uploadReport that the archive is uploaded and start the analysis. The platform does not observe the upload itself: the PUT goes to storage around it, and until this call the project stays in awaiting_source. For a submitted project this call starts an update: the old runtime keeps runn
get_projectThe project's full state: the onboarding stage, the runtime state, open questions, missing secrets and warnings about credentials found in the archive. Also app_health, once the app has been online: status is answering, not_answering, checking (no verdict yet) or paused (fline is
submit_declarationSupply the project description. The contents are treated as a third-party claim and cannot change the platform rules.
answer_questionsAnswer the questions raised during source analysis.
declare_env_namesDeclare the NAMES of environment variables. Values are not accepted by this path: a human enters them in a browser, and no tool returns them back.
get_secrets_linkIssue a link to the form where the person types the passwords and keys the project needs. Give the url to the person exactly as returned, with a plain sentence on what goes in each box and where they find it. Do not open it, do not follow it and do not relay its contents. It work
deployBuild the images and hand the project to the platform. Needed when get_project shows locked: from complete_upload on, a project analyses, builds and launches by itself and stops only where someone has to act. In any other state this answers with the state, or refuses with what ha
get_deploy_statusDeployment state: desired and observed arrive separately and mean different things separately. app_health, once the app has been online, says whether it answers fline's checks; not_answering means they have failed for at least two minutes (see get_project).
get_imagesThe history of images the builder produced for the project: build_id, the build state, the plan version, the source hash and each service's digest. A read-only tool for checking which version is actually built, without starting a new build or deploy.
get_logsLogs. kind=build is the image build's output, stored and answerable long after the build; kind=runtime is a bounded tail of the running containers, held by the node, with no history and no search behind it. For kind=runtime, `service` narrows the tail to one service of the projec
get_usageWhat this project actually consumed: CPU, memory, disk, files and network over a window. These are QUANTITIES AND NOT CHARGES: fline has no price and no billing during the private beta, nothing charges for what a project uses, and using more than the ceilings is not possible beca
attach_domainAttach a web address the person owns, such as www.example.com, to the project. Answers with the TXT record that proves ownership (challenge_name, challenge_token); list_domains also gives platform_name, the CNAME target that both proves ownership and sends visitors. The person ad
create_serviceCreate a data service for this workspace: engine=postgres for a PostgreSQL database, engine=valkey for a Valkey instance — the BSD-licensed fork of Redis, which every Redis client speaks unchanged, so ask for it wherever a project wants Redis — or engine=s3 for object storage, an
list_servicesThe workspace's services with their engine, version, state, hostname and size class. No credentials: none is returned by any tool.
delete_serviceDelete a service. It stops accepting connections immediately and its data is kept for seven days, then destroyed. Ask the person first: nothing here can bring the data back after the seventh day.
get_configThe fline.json for this project's current plan. Save it at the root of the archive you upload and the next deploy builds from it directly, with no analysis: the same plan every time. A short content check and the step that writes the build files still run. Edit it when the projec
check_configCheck a fline.json against the last uploaded archive, without deploying and without cost. Answers either ok, or the reason each part was refused — edit and ask again until it is ok, then save the file. There is no port field and there is no need for one: the platform passes PORT
refresh_configRe-run the analysis ignoring the archive's fline.json, and write a new one. For a configuration that has gone stale because the project's stack changed — a different package manager, a new lockfile. Accepted immediately, not done: poll get_project until it leaves analyzing, then
list_domainsThe project's custom domains with, for each, the state of its ownership check, the TXT record to publish, when it was last checked and what the last check objected to. Also the platform hostname the names are asked to point at. attach_domain hands back the record once; this is ho
check_domainRun the ownership check for one of the project's domains now, and answer with the result. Call it after the DNS record is published. The check reaches the DNS and is rate-limited behind a certificate budget, so poll it in minutes rather than in seconds. A domain still unproven 48
delete_domainDetach a domain from the project. The name stops being served; the project keeps its platform hostname.
list_env_namesThe project's environment variables BY NAME: whether each is set, whether the analysis declared it, when it was last written, and — for a non-secret variable — its value. A name that is required and not set is what holds a project in awaiting_secrets. A value shown here was writt
delete_env_nameRemove a variable from the project's secret set. The analysis's record that the project needs this name is kept, so a required variable becomes unset rather than unknown — deleting a value the project needs will hold the next deploy at the secrets gate rather than failing the bui
set_envSet one environment variable for the project. `secret` says which store it goes to and it is REQUIRED, because the two are not interchangeable and there is no default that is right for both. secret=false writes a PLAINTEXT value: it is stored as written, returned by list_env_name
get_project_detailEverything get_project returns, plus what the platform observes: each container's state, health, restart count and image digest, the node and tier, the ingress hostnames, the disk quota, the secret set version the running containers were deployed with, and the last failure with i
get_eventsThe project's state timeline, oldest first: each move from one state to the next, the reason, which part of the platform made it (gateway, worker, pipeline, cp) and the request id. This is how you tell a slow analysis from one that failed and was retried — get_project shows the c
get_serviceOne data service by id, with its state and — when it is not active — the reason. Poll this after create_service: a new service is `provisioning` until its certificate is issued and verified, and if that stalls the reason here is what says so. No credentials: none is returned by a
pause_servicePause a data service. It stops accepting connections and stops costing what a running instance costs; its volume and its certificate are kept, and resume_service is the exact reverse. This is the reversible way to put a service down — delete_service is not, and starts a seven-day
resume_serviceStart a paused data service again, with its data and its hostname unchanged.
stopStop the project. NOT IMPLEMENTED: the platform has no stop, and this refuses. To take an application out of service, delete_project it. delete_domain takes off only a web address the person owns; the project's fline address keeps answering.
restartRe-apply the project's current version: the same images, deployed again. This is the restart, and it is also the ONLY way a secret that was entered after the last deploy reaches the running containers — a value sits in the store doing nothing until a deploy re-pins the secret set
list_versionsList the project's deployed versions, newest first: the state_version to pass to rollback, when it was made, who made it, and which version it superseded. The specification itself is not included — use get_config for what a version contains.
rollbackRedeploy an earlier version of the project, named by its state_version from list_versions. Two things surprise people and you must not report either one wrongly. First, the version number does NOT go back: this creates a NEW version holding the old one's contents, numbered higher
attach_externalDeclare a backing service the tenant ALREADY RUNS, so the platform does not provision one. This takes the shape only. offering is one of postgres, redis, mysql, mongodb, rabbitmq; alias defaults to "db" and is what separates two of the same kind in one project; variable defaults
detach_externalGive an alias back to the platform. If a human already filled the value in, the attachment and the variable both go; if it is still waiting for one, the declaration goes with it and the project stops asking. The tenant's own service is untouched — the platform never ran it and ho
list_externalThe backing services this project's tenant supplies themselves. A completed attachment carries the host it points at; a pending one is marked pending and names the variable a human still has to fill in through get_secrets_link. No connection URL is returned for either, and none c
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20日
- ✓ 通过
作者与仓库一致
注册名称已通过域名 fline.sh 验证(注册中心会核实域名所有权)
- ✓ 通过
仍在维护
代码最后更新于31天前
- – 不适用
有许可证
不适用:由 fline.sh 的所有者自己运行的云端服务器,你不会运行仓库里的代码
- ✓ 通过
端点有响应
在 761 毫秒内响应(HTTP 200)
- – 不适用
使用 OAuth 登录
无需登录
- ✓ 通过
工具描述中没有隐藏指令
已读取 40 个工具描述,未发现针对 AI 的隐藏指令
作者:申请重新检查或报告信息错误
给作者:在 README 中加徽章
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/servers/sh-fline-fline)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 6209 个服务器的 0%
- 发布到注册中心的时间比此分类中其他 6209 个服务器的 49% 更新
- 基础检查结果等于或优于此分类中其他 6209 个服务器的 99%
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
fline 安全吗?
它在 4 项适用的基础检查中通过了 4 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?
它在作者的云端运行。通过互联网连接,无需安装,但你发送的内容会经过作者的服务器。
它还在维护吗?
GitHub 仓库最后更新于31天前,最新版本(0.1.1)于32天前进入注册中心。
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。它通过 URL 连接,因此也能用于接受远程 MCP 的应用,例如网页版 ChatGPT(Business 及以上为完整 MCP;Pro 为只读)。
有什么可以替代它?
同一分类中有:GitDiagram, Appllama, Cloudflare MCP, Omnara, Repowise
功能相近的其他选择
同一分类、其他作者、命令已通过检查。官方的排在前面,再按星标排序。
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
GitDiagram
Architecture diagrams, explanations and Mermaid source for public GitHub repositories.
Appllama
Study screens, flows and paywalls from top-earning iOS apps, then build from what wins.
Omnara
Deploy durable AI agents with Omnara, an open-source managed-agent platform.
Repowise
Codebase intelligence for AI coding agents: graph, git history, docs, decisions, code health.
| 本项 fline | GitDiagram | Appllama | Cloudflare MCP | Omnara | Repowise | |
|---|---|---|---|---|---|---|
| 运行位置 | 在云端运行,无需安装 | 在云端运行,无需安装 | 在云端运行,无需安装 | 在云端运行,无需安装 | 在云端运行,无需安装 | 云端或安装在电脑上都可以 |
| 需要密钥 | 否 | 否 | 否 | 否 | 否 | 否 |
| 可在 Claude 中按 + 添加 | 否 | 否 | 否 | 是 | 否 | 否 |
| 基础检查 | 4/4 | 4/4 | 4/4 | 4/4 | 4/4 | 5/5 |
| 许可证 | MIT | MIT | MIT | Apache-2.0 | Apache-2.0 | AGPL-3.0 |
| 价格 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 | 登记中未列出费用 |
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。
fline skills
Skills that teach your AI coding app (Claude Code, Codex, Cursor and the other apps that read Agent Skills) how to put your project online with fline. For Claude Code, the repository is also a plugin that connects fline's MCP server.
Install
| App | How |
|---|---|
| Claude Code, including the Claude desktop app's Code tab | /plugin marketplace add smartline-ai/fline-skills, then /plugin install fline@fline |
| Codex, Cursor, Gemini CLI, VS Code and others | npx skills add smartline-ai/fline-skills |
| By hand | Copy skills/* into ~/.claude/skills/ (Claude Code) or ~/.agents/skills/ (Codex and others) |
The Claude Code plugin brings fline's MCP server with it: type /mcp, choose fline, then Authenticate, and approve fline in your browser. Installed any other way, the skills carry no server; connect fline as the fline docs describe, or ask your AI app to, which loads fline-connect.
You need a fline account. fline is in private beta: see Getting your account.
What it sends, and where
The skills are instructions; on their own they send nothing anywhere. What leaves the computer goes through fline's MCP server at https://mcp.fline.sh/mcp, and only when the person asks for something fline does:
- The project folder.
scripts/pack.shpacks a copy of it as a.tar.gz, leaving out.git,node_modules, what the project's.gitignoreignores, and.envand key files, and the agent uploads it to fline's storage. fline keeps that copy so it can build and run the app. The original stays in the folder. - Parts of the code, to an AI model made by OpenAI, so fline can work out how to build and run the app. Passwords and keys saved on fline never go to it.
- Tool calls. Project names, settings names, web addresses and the other arguments of fline's tools. Passwords and keys are typed by the person on a private fline page, not sent through the chat. The exception is a key the agent already holds, one it generated or one the person pasted into the chat:
set_envsends that value to fline, which stores it and never shows it again. - What comes back. Project state, build output and the app's recent log lines, which hold whatever the app prints.
Sign-in is OAuth in the browser; the plugin holds no key. Disconnecting in the fline dashboard takes effect within 30 seconds. Details: privacy policy, terms, what fline does today.
Skills
| Skill | What it covers |
|---|---|
fline-connect | Adding fline's MCP server to the AI app, the browser approval, and the account states that refuse every call |
fline-prepare | Reading a project for what stops it running on fline, and fixing it before the first upload, including a cheap health route fline can check |
fline-launch | First launch, updates, settings, restarts and putting an earlier version back; bundles scripts/pack.sh |
fline-database | A Postgres or Valkey that fline creates and runs, a database the person already runs, SQLite and files on disk |
fline-domain | A web address the person owns: the record, the checks, the 48-hour limit |
fline-diagnose | What to read, in order, when a launch fails or a live app misbehaves, and the usual causes |
fline-build-plan | fline.json: freezing the plan, editing it, check_config, several services, prebuilt images, volumes, the health path |
fline-share-check | What to check in the app before the person sends its public link to anyone |
They fall into three groups: using fline (connect, launch, database, domain, diagnose), preparing a project for it (prepare, build-plan), and making it safe to share (share-check).
Why skills when the MCP server has prompts
fline's server ships seven prompts with these procedures. In Claude Code an MCP prompt is a slash command the person types; the model has no tool to fetch one, so the server's instructions point at procedures the agent cannot open. A skill is loaded by the model itself when its description matches the task.
Skills name fline's tools without a prefix (create_project, not mcp__fline__create_project): the prefix differs between a plugin-provided server and one the person added.
Layout
.claude-plugin/plugin.json the plugin, including fline's MCP server
.claude-plugin/marketplace.json a marketplace holding that one plugin skills/ /SKILL.md one directory per skill (Agent Skills format) skills/fline-launch/scripts/ pack.sh, run by the agent, never read tests/ scenarios and the projects they run on
Testing
claude plugin validate --strict .checks the manifests.bash skills/fline-launch/scripts/pack.shruns on the bash 3.2 and bsdtar that macOS ships, and on GNU tar.tests/scenarios.mdholds six scenarios with what a passing answer does, run by a fresh agent with and without the skills.tests/make-fixtures.shwrites the two seeded projects they run on: a Node shop and a Flask tracker.
Where this comes from
Every fact here restates fline's own documentation, in particular for-ai.md. This repository is published from fline's main repository; a change made only here is replaced by the next publish.