terraform-cloud-mcp
一个模型上下文协议(MCP)服务器,将 Claude 与 Terraform Cloud API 集成,从而允许 Claude 通过自然对话管理您的 Terraform 基础设施。
Tools
get_account_details
Get account details for a Terraform Cloud API token This endpoint shows information about the currently authenticated user or service account, useful for verifying identity, retrieving email address, and checking authentication status. It returns the same type of object as the Users API, but also includes an email address, which is hidden when viewing info about other users. API endpoint: GET /account/details Returns: Raw API response with account information from Terraform Cloud including user ID, username, email address, and authentication status See: docs/tools/account_tools.md for usage examples
list_workspaces
List workspaces in an organization. Retrieves a paginated list of all workspaces in a Terraform Cloud organization. Results can be filtered using a search string to find specific workspaces by name. Use this tool to discover existing workspaces, check workspace configurations, or find specific workspaces by partial name match. API endpoint: GET /organizations/{organization}/workspaces Args: organization: The name of the organization to list workspaces from page_number: The page number to return (default: 1) page_size: The number of items per page (default: 20, max: 100) search: Optional search string to filter workspaces by name Returns: Paginated list of workspaces with their configuration settings and metadata See: docs/tools/workspace_tools.md for usage examples
get_workspace_details
Get details for a specific workspace, identified either by ID or by org name and workspace name. Retrieves comprehensive information about a workspace including its configuration, VCS settings, execution mode, and other attributes. This is useful for checking workspace settings before operations or determining the current state of a workspace. The workspace can be identified either by its ID directly, or by the combination of organization name and workspace name. API endpoint: - GET /workspaces/{workspace_id} (when using workspace_id) - GET /organizations/{organization}/workspaces/{workspace_name} (when using org+name) Args: workspace_id: The ID of the workspace (format: "ws-xxxxxxxx") organization: The name of the organization (required if workspace_id not provided) workspace_name: The name of the workspace (required if workspace_id not provided) Returns: Comprehensive workspace details including settings, configuration and status See: docs/tools/workspace_tools.md for usage examples
force_unlock_workspace
Force unlock a workspace. This should be used with caution. Forces a workspace to unlock even when the normal unlock process isn't possible. This is typically needed when a run has orphaned a lock or when the user who locked the workspace is unavailable. This operation requires admin privileges on the workspace. WARNING: Forcing an unlock can be dangerous if the workspace is legitimately locked for active operations. Only use this when you are certain it's safe to unlock. API endpoint: POST /workspaces/{workspace_id}/actions/force-unlock Args: workspace_id: The ID of the workspace to force unlock (format: "ws-xxxxxxxx") Returns: The workspace with updated lock status and related metadata See: docs/tools/workspace_tools.md for usage examples
create_workspace
Create a new workspace in an organization. Creates a new Terraform Cloud workspace which serves as an isolated environment for managing infrastructure. Workspaces contain variables, state files, and run histories for a specific infrastructure configuration. API endpoint: POST /organizations/{organization}/workspaces Args: organization: The name of the organization name: The name to give the workspace params: Additional workspace parameters (optional): - description: Human-readable description of the workspace - execution_mode: How Terraform runs are executed (remote, local, agent) - terraform_version: Version of Terraform to use (default: latest) - working_directory: Subdirectory to use when running Terraform - vcs_repo: Version control repository configuration - auto_apply: Whether to automatically apply successful plans - file_triggers_enabled: Whether file changes trigger runs - trigger_prefixes: Directories that trigger runs when changed - trigger_patterns: Glob patterns that trigger runs when files match - allow_destroy_plan: Whether to allow destruction plans - auto_apply_run_trigger: Whether to auto-apply changes from run triggers Returns: The created workspace data including configuration, settings and metadata See: docs/tools/workspace_tools.md for usage examples
update_workspace
Update an existing workspace. Modifies the settings of a Terraform Cloud workspace. This can be used to change attributes like execution mode, VCS repository settings, description, or any other workspace configuration options. Only specified attributes will be updated; unspecified attributes remain unchanged. API endpoint: PATCH /organizations/{organization}/workspaces/{workspace_name} Args: organization: The name of the organization that owns the workspace workspace_name: The name of the workspace to update params: Workspace parameters to update (optional): - name: New name for the workspace (if renaming) - description: Human-readable description of the workspace - execution_mode: How Terraform runs are executed (remote, local, agent) - terraform_version: Version of Terraform to use - working_directory: Subdirectory to use when running Terraform - vcs_repo: Version control repository configuration (oauth-token-id, identifier) - auto_apply: Whether to automatically apply successful plans - file_triggers_enabled: Whether file changes trigger runs - trigger_prefixes: Directories that trigger runs when changed - trigger_patterns: Glob patterns that trigger runs when files match - allow_destroy_plan: Whether to allow destruction plans - auto_apply_run_trigger: Whether to auto-apply changes from run triggers Returns: The updated workspace with all current settings and configuration See: docs/tools/workspace_tools.md for usage examples
delete_workspace
Delete a workspace. Permanently deletes a Terraform Cloud workspace and all its resources including state versions, run history, and configuration versions. This action cannot be undone. WARNING: This is a destructive operation. For workspaces that have active resources, consider running a destroy plan first or use safe_delete_workspace instead. API endpoint: DELETE /organizations/{organization}/workspaces/{workspace_name} Args: organization: The name of the organization that owns the workspace workspace_name: The name of the workspace to delete Returns: Success message with no content (HTTP 204) if successful Error response with explanation if the workspace cannot be deleted See: docs/tools/workspace_tools.md for usage examples
safe_delete_workspace
Safely delete a workspace by first checking if it can be deleted. Initiates a safe delete operation which checks if the workspace has resources before deleting it. This is a safer alternative to delete_workspace as it prevents accidental deletion of workspaces with active infrastructure. The operation follows these steps: 1. Checks if the workspace has any resources 2. If no resources exist, deletes the workspace 3. If resources exist, returns an error indicating the workspace cannot be safely deleted API endpoint: POST /organizations/{organization}/workspaces/{workspace_name}/actions/safe-delete Args: organization: The name of the organization that owns the workspace workspace_name: The name of the workspace to delete Returns: Status of the safe delete operation including: - Success response if deletion was completed - Error with details if workspace has resources and cannot be safely deleted - List of resources that would be affected by deletion (if applicable) See: docs/tools/workspace_tools.md for usage examples
lock_workspace
Lock a workspace. Locks a workspace to prevent runs from being queued. This is useful when you want to prevent changes to infrastructure while performing maintenance or making manual adjustments. Locking a workspace does not affect currently running plans or applies. API endpoint: POST /workspaces/{workspace_id}/actions/lock Args: workspace_id: The ID of the workspace to lock (format: "ws-xxxxxxxx") reason: Optional reason for locking Returns: The workspace with updated lock status and related metadata See: docs/tools/workspace_tools.md for usage examples
unlock_workspace
Unlock a workspace. Removes the lock from a workspace, allowing runs to be queued. This enables normal operation of the workspace after it was previously locked. API endpoint: POST /workspaces/{workspace_id}/actions/unlock Args: workspace_id: The ID of the workspace to unlock (format: "ws-xxxxxxxx") Returns: The workspace with updated lock status and related metadata See: docs/tools/workspace_tools.md for usage examples
create_run
Create a run in a workspace Creates a new Terraform run to trigger infrastructure changes through Terraform Cloud, representing a single execution of plan and apply operations. The run queues in the workspace and executes based on the workspace's execution mode and settings. Use this to deploy new infrastructure, apply configuration changes, or destroy resources. API endpoint: POST /runs Args: workspace_id: The workspace ID to execute the run in (format: "ws-xxxxxxxx") params: Optional run configuration with: - message: Description of the run's purpose - is_destroy: Whether to destroy all resources managed by the workspace - auto_apply: Whether to auto-apply after a successful plan - refresh: Whether to refresh Terraform state before planning - refresh_only: Only refresh the state without planning changes - plan_only: Create a speculative plan without applying - allow_empty_apply: Allow applying when there are no changes - target_addrs: List of resource addresses to specifically target - replace_addrs: List of resource addresses to force replacement - variables: Run-specific variables that override workspace variables - terraform_version: Specific Terraform version to use for this run - save_plan: Save the plan for later execution - debugging_mode: Enable extended debug logging Returns: The created run details with ID, status, configuration information, workspace relationship, and links to associated resources See: docs/tools/run_tools.md for usage examples
list_runs_in_workspace
List runs in a workspace with filtering and pagination Retrieves run history for a specific workspace with options to filter by status, operation type, source, and other criteria. Useful for auditing changes, troubleshooting, or monitoring deployment history. API endpoint: GET /workspaces/{workspace_id}/runs Args: workspace_id: The workspace ID to list runs for (format: "ws-xxxxxxxx") page_number: Page number to fetch (default: 1) page_size: Number of results per page (default: 20) filter_operation: Filter by operation type filter_status: Filter by status filter_source: Filter by source filter_status_group: Filter by status group filter_timeframe: Filter by timeframe filter_agent_pool_names: Filter by agent pool names search_user: Search by VCS username search_commit: Search by commit SHA search_basic: Search across run ID, message, commit SHA, and username Returns: List of runs with metadata, status info, and pagination details See: docs/tools/run_tools.md for usage examples
list_runs_in_organization
List runs across all workspaces in an organization Retrieves run history across all workspaces in an organization with powerful filtering. Useful for organization-wide auditing, monitoring deployments across teams, or finding specific runs by commit or author. API endpoint: GET /organizations/{organization}/runs Args: organization: The organization name page_number: Page number to fetch (default: 1) page_size: Number of results per page (default: 20) filter_operation: Filter by operation type filter_status: Filter by status filter_source: Filter by source filter_status_group: Filter by status group filter_timeframe: Filter by timeframe filter_agent_pool_names: Filter by agent pool names filter_workspace_names: Filter by workspace names search_user: Search by VCS username search_commit: Search by commit SHA search_basic: Basic search across run attributes Returns: List of runs across workspaces with metadata and pagination details See: docs/tools/run_tools.md for usage examples
get_run_details
Get detailed information about a specific run Retrieves comprehensive information about a run including its current status, plan output, and relationship to other resources. Use to check run progress or results. API endpoint: GET /runs/{run_id} Args: run_id: The ID of the run to retrieve details for (format: "run-xxxxxxxx") Returns: Complete run details including status, plan, and relationships See: docs/tools/run_tools.md for usage examples
apply_run
Apply a run that is paused waiting for confirmation after a plan Confirms and executes the apply phase for a run that has completed planning and is waiting for approval. Use this when you've reviewed the plan output and want to apply the proposed changes to your infrastructure. API endpoint: POST /runs/{run_id}/actions/apply Args: run_id: The ID of the run to apply (format: "run-xxxxxxxx") comment: An optional comment explaining the reason for applying the run Returns: Run details with updated status information and confirmation of the apply action including timestamp information and any comment provided See: docs/tools/run_tools.md for usage examples
discard_run
Discard a run that is paused waiting for confirmation Cancels a run without applying its changes, typically used when the plan shows undesired changes or after reviewing and rejecting a plan. This action removes the run from the queue and unlocks the workspace for new runs. API endpoint: POST /runs/{run_id}/actions/discard Args: run_id: The ID of the run to discard (format: "run-xxxxxxxx") comment: An optional explanation for why the run was discarded Returns: Run status update with discarded state information, timestamp of the discard action, and user information See: docs/tools/run_tools.md for usage examples
cancel_run
Cancel a run that is currently planning or applying Gracefully stops an in-progress run during planning or applying phases. Use this when you need to stop a run that's taking too long, consuming too many resources, or needs to be stopped for any reason. The operation attempts to cleanly terminate the run by sending an interrupt signal. API endpoint: POST /runs/{run_id}/actions/cancel Args: run_id: The ID of the run to cancel (format: "run-xxxxxxxx") comment: An optional explanation for why the run was canceled Returns: Run status update with canceled state, timestamp of cancellation, and any provided comment in the response metadata See: docs/tools/run_tools.md for usage examples
force_cancel_run
Forcefully cancel a run immediately Immediately terminates a run that hasn't responded to a normal cancel request. Use this as a last resort when a run is stuck and not responding to regular cancellation. This action bypasses the graceful shutdown process and forces the workspace to be unlocked. API endpoint: POST /runs/{run_id}/actions/force-cancel Args: run_id: The ID of the run to force cancel (format: "run-xxxxxxxx") comment: An optional explanation for why the run was force canceled Returns: Run status update confirming forced cancellation with timestamp, user information, and workspace unlock status See: docs/tools/run_tools.md for usage examples
force_execute_run
Forcefully execute a run by canceling all prior runs Prioritizes a specific run by canceling other queued runs to unlock the workspace, equivalent to clicking "Run this plan now" in the UI. Use this when a run is stuck in the pending queue but needs immediate execution due to urgency or priority over other queued runs. API endpoint: POST /runs/{run_id}/actions/force-execute Args: run_id: The ID of the run to execute (format: "run-xxxxxxxx") Returns: Status update confirming the run has been promoted to active status, with information about which runs were canceled to allow execution See: docs/tools/run_tools.md for usage examples
get_organization_details
Get details for a specific organization Retrieves comprehensive information about an organization including settings, email contact info, and configuration defaults. API endpoint: GET /organizations/{organization} Args: organization: The organization name to retrieve details for (required) Returns: Organization details including name, email, settings and configuration See: docs/tools/organization_tools.md for usage examples
get_organization_entitlements
Show entitlement set for organization features Retrieves information about available features and capabilities based on the organization's subscription tier. API endpoint: GET /organizations/{organization}/entitlement-set Args: organization: The organization name to retrieve entitlements for (required) Returns: Entitlement set details including feature limits and subscription information See: docs/tools/organization_tools.md for usage examples
list_organizations
List organizations with filtering options Retrieves a paginated list of organizations the current user has access to, with options to search by name or email address. API endpoint: GET /organizations Args: page_number: Page number to fetch (default: 1) page_size: Number of results per page (default: 20) query: Search query to filter by name and email query_email: Search query to filter by email only query_name: Search query to filter by name only Returns: List of organizations with metadata and pagination information See: docs/tools/organization_tools.md for usage examples
create_organization
Create a new organization in Terraform Cloud Creates a new organization with the given name and email, allowing workspaces and teams to be created within it. This is the first step in setting up a new environment in Terraform Cloud. API endpoint: POST /organizations Args: name: The name of the organization (required) email: Admin email address (required) params: Additional organization settings: - collaborator_auth_policy: Authentication policy (password or two_factor_mandatory) - session_timeout: Session timeout after inactivity in minutes - session_remember: Session total expiration time in minutes - cost_estimation_enabled: Whether to enable cost estimation for workspaces - default_execution_mode: Default workspace execution mode (remote, local, agent) - aggregated_commit_status_enabled: Whether to aggregate VCS status updates - speculative_plan_management_enabled: Whether to auto-cancel unused speculative plans - assessments_enforced: Whether to enforce health assessments for all workspaces - allow_force_delete_workspaces: Whether to allow deleting workspaces with resources - default_agent_pool_id: Default agent pool ID (required when using agent mode) Returns: The created organization details including ID and created timestamp See: docs/tools/organization_tools.md for usage examples
update_organization
Update an existing organization in Terraform Cloud Modifies organization settings such as email contact, authentication policy, or other configuration options. Only specified attributes will be updated. API endpoint: PATCH /organizations/{organization} Args: organization: The name of the organization to update (required) params: Organization parameters to update: - email: Admin email address for the organization - collaborator_auth_policy: Authentication policy (password or two_factor_mandatory) - session_timeout: Session timeout after inactivity in minutes - session_remember: Session total expiration time in minutes - cost_estimation_enabled: Whether to enable cost estimation for workspaces - default_execution_mode: Default workspace execution mode (remote, local, agent) - aggregated_commit_status_enabled: Whether to aggregate VCS status updates - speculative_plan_management_enabled: Whether to auto-cancel unused speculative plans - assessments_enforced: Whether to enforce health assessments for all workspaces - allow_force_delete_workspaces: Whether to allow deleting workspaces with resources Returns: The updated organization with all current settings See: docs/tools/organization_tools.md for usage examples
delete_organization
Delete an organization from Terraform Cloud Permanently removes an organization including all its workspaces, teams, and resources. This action cannot be undone. Organization names are globally unique and cannot be recreated with the same name later. API endpoint: DELETE /organizations/{organization} Args: organization: The name of the organization to delete (required) Returns: Success confirmation (HTTP 204 No Content) or error details See: docs/tools/organization_tools.md for usage examples
README
Terraform Cloud MCP 服务器
一个模型上下文协议 (MCP) 服务器,它将 AI 助手与 Terraform Cloud API 集成,允许您通过自然对话管理您的基础设施。该服务器使用 Pydantic 模型构建,并围绕特定领域的模块进行组织,与任何支持 MCP 的平台兼容,包括 Claude、Claude Code CLI、Claude Desktop、Cursor、Copilot Studio 等。
功能
- 账户管理: 获取已验证用户或服务账户的账户详细信息。
- 工作区管理: 创建、读取、更新、删除、锁定/解锁工作区。
- 运行管理: 创建运行、列出运行、获取运行详细信息、应用/丢弃/取消运行。
- 计划管理: 检索计划详细信息和 JSON 执行输出,并具有高级 HTTP 重定向处理。
- 应用管理: 获取应用详细信息并从失败的状态上传中恢复。
- 组织管理: 列出、创建、更新、删除组织,并查看组织权利。
- 未来功能: 变量管理、状态版本等。
快速开始
前提条件
- Python 3.12+
- MCP (包括 FastMCP 和开发工具)
uv包管理器 (推荐) 或pip- Terraform Cloud API 令牌
安装
# 克隆仓库
git clone https://github.com/severity1/terraform-cloud-mcp.git
cd terraform-cloud-mcp
# 创建虚拟环境并激活它
uv venv
source .venv/bin/activate
# 安装包
uv pip install .
添加到 Claude 环境
添加到 Claude Code CLI
# 使用您的 Terraform Cloud 令牌添加到 Claude Code
claude mcp add -e TFC_TOKEN=YOUR_TF_TOKEN -s user terraform-cloud-mcp -- "terraform-cloud-mcp"
添加到 Claude Desktop
创建一个 claude_desktop_config.json 配置文件:
- mac: ~/Library/Application Support/Claude/claude_desktop_config.json
- win: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"terraform-cloud-mcp": {
"command": "/path/to/uv", # 通过运行 `which uv` 获取
"args": [
"--directory",
"/path/to/your/terraform-cloud-mcp", # 此项目的完整路径
"run",
"terraform-cloud-mcp"
],
"env": {
"TFC_TOKEN": "my token..." # 替换为实际令牌
}
}
}
}
将 your_terraform_cloud_token 替换为您的实际 Terraform Cloud API 令牌。
其他 MCP 兼容平台
对于其他平台(如 Cursor、Copilot Studio 或 Glama),请按照其平台特定的说明添加 MCP 服务器。大多数平台需要:
- 服务器路径或启动服务器的命令。
- Terraform Cloud API 令牌的环境变量。
- 配置为在需要时自动启动服务器。
可用工具
账户工具
get_account_details(): 获取已验证用户或服务账户的账户信息。
工作区管理工具
列表 & 搜索
list_workspaces(organization, page_number, page_size, search): 列出和筛选工作区。get_workspace_details(workspace_id, organization, workspace_name): 获取有关特定工作区的详细信息。
创建 & 更新
create_workspace(organization, name, params): 使用可选参数创建一个新的工作区。update_workspace(organization, workspace_name, params): 更新现有工作区的配置。
删除
delete_workspace(organization, workspace_name): 删除工作区及其所有内容。safe_delete_workspace(organization, workspace_name): 仅当工作区未管理任何资源时才删除。
锁定 & 解锁
lock_workspace(workspace_id, reason): 锁定工作区以防止运行。unlock_workspace(workspace_id): 解锁工作区以允许运行。force_unlock_workspace(workspace_id): 强制解锁由其他用户锁定的工作区。
<!-- Future implementation: Data Retention
set_data_retention_policy(workspace_id, days): Set a data retention policy.get_data_retention_policy(workspace_id): Get the current data retention policy.delete_data_retention_policy(workspace_id): Delete the data retention policy. -->
运行管理工具
create_run(workspace_id, params): 使用工作区的 ID 在工作区中创建并排队一个 Terraform 运行。list_runs_in_workspace(workspace_id, ...): 使用工作区的 ID 列出和筛选特定工作区中的运行。list_runs_in_organization(organization, ...): 列出和筛选整个组织中的运行。get_run_details(run_id): 获取有关特定运行的详细信息。apply_run(run_id, comment): 应用等待确认的运行。discard_run(run_id, comment): 丢弃等待确认的运行。cancel_run(run_id, comment): 取消当前正在计划或应用的运行。force_cancel_run(run_id, comment): 立即强制取消运行。force_execute_run(run_id): 通过取消先前的运行来强制执行挂起的运行。
计划管理工具
get_plan_details(plan_id): 获取有关特定计划的详细信息。get_plan_json_output(plan_id): 检索特定计划的 JSON 执行计划,并具有正确的重定向处理。get_run_plan_json_output(run_id): 检索运行中的 JSON 执行计划,并具有正确的重定向处理。
应用管理工具
get_apply_details(apply_id): 获取有关特定应用的详细信息。get_errored_state(apply_id): 检索来自失败应用的错误状态以进行恢复。
组织管理工具
get_organization_details(organization): 获取有关特定组织的详细信息。get_organization_entitlements(organization): 显示组织功能的权利集。list_organizations(page_number, page_size, query, query_email, query_name): 列出和筛选组织。create_organization(name, email, params): 使用可选参数创建一个新的组织。update_organization(organization, params): 更新现有组织的设置。delete_organization(organization): 删除组织及其所有内容。
开发指南
有关详细的开发指导,包括代码标准、Pydantic 模式和贡献工作流程,请参阅我们的 开发文档。
快速开发设置
# 克隆仓库
git clone https://github.com/severity1/terraform-cloud-mcp.git
cd terraform-cloud-mcp
# 创建虚拟环境并激活它
uv venv
source .venv/bin/activate # 在 Windows 上: .venv\Scripts\activate
# 以开发模式安装,并包含开发依赖项
uv pip install -e .
uv pip install black mypy pydantic ruff
基本开发命令
# 在开发模式下运行服务器
mcp dev terraform_cloud_mcp/server.py
# 运行测试和质量检查
uv run -m mypy .
uv run -m ruff check .
uv run -m black .
有关代码组织、架构、开发工作流程和代码质量准则的详细信息,请参阅 docs/DEVELOPMENT.md。
文档
代码库包含全面的文档:
- 代码注释: 侧重于解释实现决策背后的“原因”
- 文档字符串: 所有公共函数和类都包含详细的文档字符串
- 示例文件:
docs/目录包含每个领域的详细示例:docs/DEVELOPMENT.md: 开发标准和编码指南docs/CONTRIBUTING.md: 贡献项目的指南docs/models/: 所有模型类型的使用示例docs/tools/: 每个工具的详细使用示例docs/conversations/: 与 API 的示例对话流程
故障排除
- 检查服务器日志(默认情况下启用调试日志记录)
- 使用 MCP Inspector (http://localhost:5173) 进行调试
- 调试日志记录已在
server.py中启用:import logging logging.basicConfig(level=logging.DEBUG)
贡献
欢迎贡献!如果您想为该项目做出贡献,请打开一个 issue 或 pull request。
请参阅我们的 贡献指南,以获取有关如何入门、代码质量标准和 pull request 流程的详细说明。
推荐服务器
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
AIO-MCP Server
🚀 集成了 AI 搜索、RAG 和多服务(GitLab/Jira/Confluence/YouTube)的一体化 MCP 服务器,旨在增强 AI 驱动的开发工作流程。来自 Folk。
https://github.com/Streen9/react-mcp
react-mcp 与 Claude Desktop 集成,能够根据用户提示创建和修改 React 应用程序。
MCP Atlassian
适用于 Atlassian Cloud 产品(Confluence 和 Jira)的 Model Context Protocol (MCP) 服务器。此集成专为 Atlassian Cloud 实例设计,不支持 Atlassian Server 或 Data Center 部署。
any-chat-completions-mcp
将 Claude 与任何 OpenAI SDK 兼容的聊天完成 API 集成 - OpenAI、Perplexity、Groq、xAI、PyroPrompts 等。
Exa MCP Server
一个模型上下文协议服务器,它使像 Claude 这样的人工智能助手能够以安全和受控的方式,使用 Exa AI 搜索 API 执行实时网络搜索。
MySQL MCP Server
允许人工智能助手通过受控界面列出表格、读取数据和执行 SQL 查询,从而使数据库探索和分析更安全、更有条理。