模型上下文协议 (MCP) 服务器用于向 Claude Desktop 等 LLM 工具提供 Salesforce CLI 功能。
该 MCP 服务器包装了 Salesforce CLI（ sf ）命令行工具，并将其命令公开为 MCP 工具和资源，从而允许 LLM 驱动的代理执行以下操作：
查看有关 Salesforce CLI 主题和命令的帮助信息
使用适当的参数执行 Salesforce CLI 命令
在 AI 工作流中利用 Salesforce CLI 功能
Related MCP server: Salesforce CLI MCP Server
Node.js 18+ 和 npm
Salesforce CLI（ sf ）已安装并配置
在 CLI 中配置的 Salesforce 组织凭据
# Clone the repository
git clone <repository-url>
cd sfMcp
# Install dependencies
npm install# Basic usage
npm start
# With project roots
npm start /path/to/project1 /path/to/project2
# or using the convenience script
npm run with-roots /path/to/project1 /path/to/project2
# As an npx package with roots
npx -y codefriar/sf-mcp /path/to/project1 /path/to/project2MCP 服务器使用 stdio 传输，可与 MCP 客户端（如MCP Inspector或 Claude Desktop）一起使用。
要在 Claude Desktop 的.claude.json配置中配置此 MCP：
{
"tools": {
"salesforce": {
"command": "/path/to/node",
"args": [
"/path/to/sf-mcp/build/index.js",
"/path/to/project1",
"/path/to/project2"
]
}
}
}直接使用 npm 包：
{
"tools": {
"salesforce": {
"command": "/path/to/npx",
"args": [
"-y",
"codefriar/sf-mcp",
"/path/to/project1",
"/path/to/project2"
]
}
}
}# Watch mode (recompiles on file changes)
npm run dev
# In another terminal
npm start [optional project roots...]此 MCP 服务器提供 Salesforce CLI 命令作为 MCP 工具。它会自动发现并注册 Salesforce CLI 中所有可用的命令，并专门实现最常用的命令。
sf_version - 获取 Salesforce CLI 版本信息
sf_help - 获取 Salesforce CLI 命令的帮助信息
sf_cache_clear - 清除命令发现缓存
sf_cache_refresh - 刷新命令发现缓存
对于需要 Salesforce 项目上下文的命令（例如部署），您必须指定项目目录。MCP 支持多个项目目录（根目录），类似于文件系统 MCP。
方法 1：通过命令行参数
# Start the MCP with project roots
npm start /path/to/project1 /path/to/project2
# or
npx -y codefriar/sf-mcp /path/to/project1 /path/to/project2当以这种方式配置时，根将自动命名为root1 、 root2等，其中第一个设置为默认值。
方法 2：使用 MCP 工具
sf_set_project_directory - 设置用于命令的 Salesforce 项目目录
参数：
directory - 包含 sfdx-project.json 文件的目录路径
name -（可选）此项目根目录的名称
description - （可选）此项目根目录的描述
isDefault - （可选）将此根设置为命令执行的默认根
sf_list_roots - 列出所有已配置的项目根目录
sf_detect_project_directory - 尝试从用户消息中检测项目目录
使用示例：
# Set project directory with a name
sf_set_project_directory --directory=/path/to/your/sfdx/project --name=project1 --isDefault=true
# List all configured roots
sf_list_roots
# Or include in your message:
"Please deploy the apex code from the project in /path/to/your/sfdx/project to my scratch org"方法 3：Claude 桌面配置按照如下所述在.claude.json中配置项目根。
您可以在特定的项目根目录中执行命令：
# Using resource URI
sf://roots/project1/commands/project deploy start --sourcedir=force-app
# Using rootName parameter
sf_project_deploy_start --sourcedir=force-app --rootName=project1必须为部署、源代码检索和其他特定于项目的操作等命令指定项目目录。如果配置了多个根目录，则除非另有说明，否则将使用默认根目录。
以下命令是专门实现的，并保证可以正常工作：
sf_org_list - 列出 Salesforce 组织
参数： json ， verbose
sf_auth_list_orgs - 列出经过身份验证的 Salesforce 组织
参数： json ， verbose
sf_org_display - 显示有关组织的详细信息
参数： targetusername ， json
sf_org_open - 在浏览器中打开一个组织
参数： targetusername 、 path 、 urlonly
sf_apex_run - 运行匿名 Apex 代码
参数： targetusername 、 file 、 apexcode 、 json
sf_apex_test_run - 运行 Apex 测试
参数： targetusername 、 testnames 、 suitenames 、 classnames 、 json
sf_data_query - 执行 SOQL 查询
参数： targetusername 、 query 、 json
sf_schema_list_objects - 列出组织中的 sObjects
参数： targetusername ， json
sf_schema_describe - 描述 Salesforce 对象
参数： targetusername 、 sobject 、 json
sf_project_deploy_start - 将源部署到组织
参数： targetusername 、 sourcedir 、 json 、 wait
服务器发现所有可用的 Salesforce CLI 命令并将它们注册为格式为sf_<topic>_<command>工具。
例如：
sf_apex_run - 运行匿名 Apex 代码
sf_data_query - 执行 SOQL 查询
对于嵌套主题命令，工具名称包含带下划线的完整路径：
sf_apex_log_get - 获取 apex 日志
sf_org_login_web - 使用 web flow 登录到组织
服务器还会尽可能为常见的嵌套命令创建简化的别名：
sf_get作为sf_apex_log_get的别名
sf_web作为sf_org_login_web的别名
可用的命令取决于安装的 Salesforce CLI 插件。
**注意：**命令发现会被缓存以提高启动性能。如果您安装了新的 SF CLI 插件，请使用
sf_cache_refresh工具更新缓存，然后重新启动服务器。
以下资源提供有关 Salesforce CLI 的文档：
sf://help主要 CLI 文档
sf://topics/{topic}/help - 主题帮助文档
sf://commands/{command}/help - 命令帮助文档
sf://topics/{topic}/commands/{command}/help - 主题命令帮助文档
sf://version版本信息
sf://roots列出所有已配置的项目根目录
sf://roots/{root}/commands/{command} - 在特定项目根目录中执行命令
在启动时，服务器会检查缓存的命令列表（存储在~/.sf-mcp/command-cache.json中）
如果存在有效的缓存，则使用它来注册命令；否则，将动态发现命令
在发现期间，服务器查询sf commands --json以获取可用命令的完整列表
命令元数据（包括参数和描述）直接从 JSON 输出中提取
所有命令均注册为具有适当参数模式的 MCP 工具
资源已注册用于帮助文档
当调用工具时，将执行相应的 Salesforce CLI 命令
对于需要 Salesforce 项目上下文的命令：
服务器通过sf_set_project_directory检查是否已配置任何项目根目录
如果配置了多个根，则使用默认根，除非指定了特定的根
如果没有设置根目录，服务器将提示用户指定项目目录
命令在适当的项目目录中执行，确保正确的上下文
用户可以根据需要添加或在多个项目根之间切换
项目特定的命令（例如部署、检索等）将自动在相应的项目目录中运行。对于不需要项目上下文的命令，工作目录无关紧要。
您可以通过以下方式在特定项目根目录中执行命令：
使用资源 URI： sf://roots/{rootName}/commands/{command}
为命令工具提供rootName参数（内部实现细节）
使用sf_set_project_directory --isDefault=true将特定根目录设置为默认根目录
为了提高启动性能，MCP 服务器会缓存发现的命令：
缓存存储在~/.sf-mcp/command-cache.json
它包括所有主题、命令、参数和描述
缓存具有验证时间戳和 SF CLI 版本检查
默认情况下，缓存会在 7 天后过期
安装新的 Salesforce CLI 插件时，使用sf_cache_refresh更新缓存
服务器首次运行会执行完整的命令发现，这可能需要一些时间。如果您遇到任何命令缺失或缓存问题：
停止 MCP 服务器（如果正在运行）
手动删除缓存文件： rm ~/.sf-mcp/command-cache.json
再次启动服务器： npm start
这将强制使用官方 CLI 元数据重新发现所有命令。
如果仍然缺少特定命令，或者您已安装了新的 SF CLI 插件：
使用 Claude Desktop 的sf_cache_refresh工具
停止并重新启动 MCP 服务器
Salesforce CLI 具有多层级的命令结构。MCP 服务器通过以下方式处理这些嵌套命令：
将冒号分隔的路径转换为下划线格式（ apex:log:get → sf_apex_log_get ）
尽可能为常见的深度命令提供别名（ sf_get用于sf_apex_log_get ）
在工具名称中保留完整的命令层次结构
使用sf commands --json
嵌套主题命令尽可能注册两次 - 一次使用完整的层次结构名称，一次使用简化的别名，以便更容易发现和使用它们。
国际学习中心
Clear the cached SF command metadata to force a refresh
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the burden of behavioral disclosure. It indicates a mutation action ('Clear') that likely affects system state, but does not detail permissions, side effects, or error handling. It adds value by specifying the purpose (force refresh) but lacks depth on behavioral traits like reversibility or impact.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that directly states the action and outcome with no wasted words. It is front-loaded and appropriately sized for a simple tool, making it easy to understand quickly.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (simple mutation with no parameters) and lack of annotations or output schema, the description is minimally adequate. It explains what the tool does but does not cover behavioral aspects like effects or usage context in depth. For a mutation tool, more disclosure would be beneficial, but it meets basic requirements.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 0 parameters with 100% coverage, so no parameter documentation is needed. The description does not add parameter details, which is appropriate. Baseline is 4 for zero parameters, as the description focuses on tool purpose without unnecessary parameter repetition.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('Clear') and the target ('cached SF command metadata'), with the specific outcome of forcing a refresh. It distinguishes from siblings like sf_cache_refresh by emphasizing clearing rather than refreshing, though the distinction could be more explicit. No tautology or vagueness is present.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage when a refresh of metadata is needed, but does not explicitly state when to use this tool versus alternatives like sf_cache_refresh or other siblings. It provides some context (force a refresh) but lacks explicit guidance on prerequisites, exclusions, or comparisons with other tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Refresh the SF command cache by re-scanning all available commands
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It states the tool refreshes the cache by re-scanning commands, implying a read/write operation that updates cache state, but doesn't specify permissions needed, side effects, performance impact, or error handling. For a tool that likely modifies cache data, this is insufficient behavioral context.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that directly states the tool's action without unnecessary words. It is front-loaded with the core purpose ('Refresh the SF command cache') and adds clarifying detail ('by re-scanning all available commands'). Every part of the sentence contributes meaning, making it highly concise and well-structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has no parameters, no annotations, and no output schema, the description is minimal. While it states what the tool does, it lacks context about why or when to use it, behavioral traits like side effects or performance, and what the output might be. For a cache operation tool, this leaves significant gaps in understanding its role and impact.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has 0 parameters, and schema description coverage is 100% (as there are no parameters to describe). The description doesn't need to add parameter semantics, so it meets the baseline of 4 for zero-parameter tools. No additional value is required beyond stating the tool's purpose.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb ('Refresh') and resource ('SF command cache'), and specifies the action ('by re-scanning all available commands'). It distinguishes from sibling 'sf_cache_clear' by indicating a refresh rather than a clear operation. However, it doesn't explicitly differentiate from other siblings like project directory tools, keeping it at 4 rather than 5.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives like 'sf_cache_clear' or other cache-related operations. It lacks context about prerequisites, timing, or scenarios where this refresh is necessary versus other actions. This leaves the agent without usage direction.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Get instructions for setting up Salesforce project directories for command execution
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It states the tool 'Get instructions,' implying a read-only operation that returns guidance, but it doesn't clarify if this requires specific permissions, what format the instructions are in, or if there are any side effects. For a tool with zero annotation coverage, this leaves significant behavioral gaps.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, clear sentence: 'Get instructions for setting up Salesforce project directories for command execution.' It is front-loaded with the main purpose, has no unnecessary words, and efficiently conveys the tool's intent without waste.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has 0 parameters, no annotations, and no output schema, the description is minimally adequate. It states what the tool does but lacks details on the instruction format, prerequisites, or how it relates to sibling tools. For a tool that likely provides setup guidance, more context on the output or usage scenarios would improve completeness.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has 0 parameters, and schema description coverage is 100%, so there are no parameters to document. The description doesn't need to add parameter semantics beyond what the schema provides. A baseline score of 4 is appropriate as the description doesn't introduce confusion about parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Get instructions for setting up Salesforce project directories for command execution.' It specifies the action ('Get instructions') and resource ('Salesforce project directories'), though it doesn't explicitly distinguish it from sibling tools like sf_set_project_directory. The purpose is clear but lacks sibling differentiation.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. It doesn't mention prerequisites, timing, or relationships with sibling tools such as sf_set_project_directory (which might actually set the directory) or sf_list_roots (which might list available directories). Without such context, the agent lacks clear usage direction.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List all configured Salesforce project directories and their metadata
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It states what the tool does but doesn't cover important aspects like whether it's read-only, requires authentication, has rate limits, or what the output format looks like. This leaves significant gaps in understanding the tool's behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that directly states the tool's purpose without any unnecessary words. It's appropriately sized and front-loaded, making it easy for an agent to parse quickly.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has no parameters and no output schema, the description provides basic purpose information but lacks important context about behavior, output format, and usage guidelines. For a tool with no annotations and no output schema, more completeness would be expected to help the agent understand what to expect from the operation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has 0 parameters with 100% schema description coverage, so the schema already fully documents the input requirements. The description appropriately doesn't add parameter information beyond what's in the schema, maintaining a baseline score of 4 for tools with no parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('List') and the resource ('all configured Salesforce project directories and their metadata'), providing a specific verb+resource combination. However, it doesn't explicitly differentiate from sibling tools like sf_detect_project_directory or sf_set_project_directory, which prevents a perfect score.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. There's no mention of prerequisites, timing, or comparison to siblings like sf_cache_clear or sf_detect_project_directory, leaving the agent without contextual usage instructions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Set a Salesforce project directory for command execution context
| Name | Required | Description | Default |
|---|---|---|---|
| directory | Yes | The absolute path to a directory containing an sfdx-project.json file | |
| name | No | Optional name for this project root | |
| description | No | Optional description for this project root | |
| isDefault | No | Set this root as the default for command execution |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It mentions setting a directory for 'command execution context,' which implies configuration/mutation, but doesn't specify whether this persists across sessions, requires specific permissions, or has side effects. For a mutation tool with zero annotation coverage, this is a significant gap in transparency.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, clear sentence that directly states the tool's purpose without unnecessary words. It's front-loaded and efficient, making it easy for an agent to parse quickly.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has 4 parameters, no annotations, and no output schema, the description is minimally adequate but incomplete. It covers the basic purpose but lacks details on behavioral traits, usage context, and output expectations, which are crucial for a mutation tool in a set of related Salesforce commands.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema description coverage is 100%, meaning all parameters are documented in the schema itself. The description adds no additional semantic information about parameters beyond what's in the schema, such as usage examples or constraints. This meets the baseline for high schema coverage but doesn't provide extra value.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('Set') and the resource ('Salesforce project directory for command execution context'), making the purpose understandable. However, it doesn't explicitly differentiate from sibling tools like sf_detect_project_directory or sf_list_roots, which prevents a perfect score.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. There's no mention of prerequisites (e.g., needing an sfdx-project.json file), when not to use it, or how it relates to sibling tools like sf_detect_project_directory. This leaves the agent with minimal context for tool selection.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.0.0sf_cache_clear1 field changedInput schema / additionalPropertiesRemoved value: -falsesf_cache_refresh1 field changedInput schema / additionalPropertiesRemoved value: -falsesf_detect_project_directory1 field changedInput schema / additionalPropertiesRemoved value: -falsesf_list_roots1 field changedInput schema / additionalPropertiesRemoved value: -falsesf_cache_clearsf_cache_refreshsf_detect_project_directorysf_list_rootssf_set_project_directoryScored across 5 tools
Each tool has a clearly distinct purpose with no overlap: cache_clear and cache_refresh handle metadata caching, detect_project_directory provides setup instructions, list_roots enumerates existing directories, and set_project_directory establishes execution context. The descriptions reinforce these boundaries, making misselection unlikely.
All tools follow a consistent 'sf_verb_noun' pattern with snake_case throughout (e.g., sf_cache_clear, sf_detect_project_directory). This predictable naming scheme makes the tool set easy to navigate and understand at a glance.
Five tools is reasonable for a Salesforce CLI server focused on project directory management and cache operations. It's slightly lean but covers core setup and maintenance tasks without feeling bloated or incomplete for its apparent scope.
The tool set covers project directory setup, listing, and context setting, plus cache management, but lacks direct Salesforce CLI command execution tools (e.g., running queries or deploying metadata). This creates a notable gap for agents needing to perform actual Salesforce operations, though the provided tools support preparatory workflows.
Plan Salesforce deploys, open pull requests and trigger pipelines from your AI client.
Provides capabilities that let LLM agents perform a range of infrastructure management tasks.
Run SOQL queries to explore and retrieve Salesforce data. Access accounts, contacts, opportunities…
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
