Dingtalk Ai Table

v0.6.0

钉钉 AI 表格(多维表)操作技能。使用 mcporter CLI 连接钉钉官方新版 AI 表格 MCP server,基于 baseId / tableId / fieldId / recordId 体系执行 Base、Table、Field、Record 的查询与增删改。适用于创建 AI 表格、搜索表格、读取...

10· 3.6k·35 current·37 all-time
byMarila Wang@aliramw

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for aliramw/dingtalk-ai-table.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Dingtalk Ai Table" (aliramw/dingtalk-ai-table) from ClawHub.
Skill page: https://clawhub.ai/aliramw/dingtalk-ai-table
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Required env vars: DINGTALK_MCP_URL, OPENCLAW_WORKSPACE
Required binaries: mcporter, python3
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install dingtalk-ai-table

ClawHub CLI

Package manager switcher

npx clawhub@latest install dingtalk-ai-table
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name/description, required binaries (mcporter, python3), required env vars (DINGTALK_MCP_URL, OPENCLAW_WORKSPACE), and included scripts all match a tool for operating the DingTalk AI Table MCP server. There are no unrelated credentials or binaries requested.
Instruction Scope
Runtime instructions call mcporter and Python scripts, inspect MCP schema, and read/write files under the declared OPENCLAW_WORKSPACE. The one-time schema check writes a small cache file into the workspace; this is consistent with the stated 'workspace sandbox' model. Minor documentation inconsistency: examples sometimes pass the raw URL to mcporter and sometimes refer to the registered name; this is a usability/documentation issue, not a security mismatch.
Install Mechanism
No install spec is provided (instruction-only plus utility scripts). Dependencies are standard (npm-installed mcporter or system mcporter CLI); there are no remote downloads or obscure installers in the package.
Credentials
The skill only requests DINGTALK_MCP_URL (the MCP server Streamable HTTP URL containing an access token) and OPENCLAW_WORKSPACE. Both are justified by the described functionality. Minor point: the schema-check script computes an md5 of the full URL and uses it for a cache filename — this stores a derivative of the credential in a workspace path (md5 is one-way but unsalted), so choose a dedicated workspace directory and protect it.
Persistence & Privilege
always is false and the skill does not request elevated/global persistent privileges. It writes small cache files and reads/writes user files only under OPENCLAW_WORKSPACE (or current directory if unset); that behavior is declared in SKILL.md and present in scripts.
Assessment
This skill appears to do what it says: it uses mcporter and Python scripts to operate DingTalk AI tables and requires the MCP Streamable HTTP URL (DINGTALK_MCP_URL) plus a local workspace root for file I/O. Before installing: (1) review and protect the DINGTALK_MCP_URL (it includes an access token) and prefer storing it in mcporter config rather than exposing it broadly; (2) set OPENCLAW_WORKSPACE to a dedicated directory you control (the skill will create a .cache/ entry and write a schema-check file whose filename is derived from an MD5 of the URL); (3) inspect the included scripts if you need extra assurance (they call mcporter and perform JSON parsing, but contain explicit input validation and path sandboxing); (4) ensure mcporter is the official CLI and keep it up-to-date. If you need stricter limits, run the scripts in a confined/test environment first.

Like a lobster shell, security has layers — review code before you run it.

Runtime requirements

Binsmcporter, python3
EnvDINGTALK_MCP_URL, OPENCLAW_WORKSPACE
Primary envDINGTALK_MCP_URL
ai-tablevk97686b34z1b83vzjnpkw5xmqx83yk3ndingtalkvk97686b34z1b83vzjnpkw5xmqx83yk3nlatestvk97686b34z1b83vzjnpkw5xmqx83yk3nmcpvk97686b34z1b83vzjnpkw5xmqx83yk3nopenclawvk97686b34z1b83vzjnpkw5xmqx83yk3nskillvk97686b34z1b83vzjnpkw5xmqx83yk3n
3.6kdownloads
10stars
29versions
Updated 4w ago
v0.6.0
MIT-0

钉钉 AI 表格操作(新版 MCP)

🚀 5 分钟快速开始

1️⃣ 列出我的表格

mcporter call '<DINGTALK_MCP_URL>' .list_bases limit=5

2️⃣ 创建新表格

mcporter call '<DINGTALK_MCP_URL>' .create_base baseName='我的项目'

3️⃣ 添加记录

mcporter call '<DINGTALK_MCP_URL>' .create_records \
  --args '{"baseId":"base_xxx","tableId":"tbl_xxx","records":[{"cells":{"fld_name":"张三"}}]}'

4️⃣ 查询记录

mcporter call '<DINGTALK_MCP_URL>' .query_records \
  --args '{"baseId":"base_xxx","tableId":"tbl_xxx","limit":10}'

5️⃣ 批量导入

python3 scripts/import_records.py base_xxx tbl_xxx data.csv

核心概念

新版 MCP schema 工作:

  • Base:baseId
  • Table:tableId
  • Field:fieldId
  • Record:recordId

不要再用旧版 dentryUuid / sheetIdOrName / fieldIdOrName

推荐使用 mcporter 0.8.1 及以上版本。

输出模式兼容说明:

  • mcporter 0.8.1+ 可直接调用
  • 更低版本需要显式加 --output text
  • AI 表格 MCP 无论使用哪种模式,返回体本身都是标准 JSON;差异主要在 mcporter 的输出处理方式

版本守门规则(每个 MCP Server 地址只强制检查一次)

在真正开始任何 AI 表格操作前,必须先检查当前 mcporter 注册的 dingtalk-ai-table MCP server 实际返回的 tools schema。但这个检查不该每次都重复做;同一个 MCP Server 地址只需要强制检查一次。

一次性检查策略

  1. 先读取当前 mcporterdingtalk-ai-table 对应的 MCP Server 地址。
  2. 用这个地址生成一个本地检查标记(例如基于完整 URL 或其 hash)。
  3. 在工作区保存检查结果,例如放到:
~/.openclaw/workspace/.cache/dingtalk-ai-table/

建议文件名模式:

schema-check-<url-hash>.json
  1. 如果当前地址对应的检查标记已经存在,并且结果是“已确认新版 schema”,则跳过重复检查,直接继续后续 AI 表格操作。
  2. 只有在以下情况才重新强制检查:
    • 第一次运行,没有检查标记
    • mcporter 里的 MCP Server 地址变了
    • 之前检查结果是旧版 schema / 检查失败
    • 用户明确要求重新验证

强制检查时执行

mcporter list dingtalk-ai-table --schema

判断标准

如果返回的 tools 仍然是旧版这一套,例如出现:

  • get_root_node_of_my_document
  • create_base_app
  • list_base_tables
  • add_base_record
  • search_base_record
  • list_base_field

或者整体仍然基于:

  • dentryUuid
  • sheetIdOrName
  • fieldIdOrName

那么说明:虽然 skill 文件已经是新版,但 mcporter 里注册的 MCP server 地址还是旧的,不能继续操作。

遇到旧版 schema 时的强制提示

此时必须明确提示用户:

  1. 打开这个页面: https://mcp.dingtalk.com/#/detail?mcpId=9555&detailType=marketMcpDetail
  2. 点击右侧 「获取 MCP Server 配置」 按钮
  3. 复制新的 MCP Server 地址
  4. 用新的地址替换 mcporter 里已经注册的 dingtalk-ai-table 地址
  5. 替换完成后,再重新执行:
mcporter list dingtalk-ai-table --schema

只有当返回的 tools 已经变成新版 schema,例如出现:

  • list_bases
  • get_base
  • get_tables
  • get_fields
  • query_records
  • create_records
  • update_records
  • delete_records
  • prepare_attachment_upload

才允许继续真正的 AI 表格操作。

通过检查后的处理

一旦确认当前 MCP Server 地址返回的是新版 schema,就把结果写入本地检查标记。后续只要 mcporter 里的 dingtalk-ai-table 地址没变,就不要再重复做这一步守门检查。

用户提示文案(可直接复用)

当前 mcporter 里注册的 dingtalk-ai-table 还是旧版 MCP schema,暂时不能按新版技能操作。
请打开 https://mcp.dingtalk.com/#/detail?mcpId=9555&detailType=marketMcpDetail ,点击右侧“获取 MCP Server 配置”按钮,复制新的 MCP Server 地址,并替换 mcporter 里已注册的 dingtalk-ai-table 地址。替换后重新检查 schema,确认出现 list_bases / get_base / create_records 等新版 tools 后,再继续操作 AI 表格。

前置要求

安装 mcporter CLI

npm install -g mcporter
# 或
bun install -g mcporter

验证:

mcporter --version

配置 MCP Server

在钉钉 MCP 广场 https://mcp.dingtalk.com/#/detail?mcpId=9555&detailType=marketMcpDetail 获取新版钉钉 AI 表格 MCP 的 Streamable HTTP URL

方式一:直接配置到 mcporter

mcporter config add dingtalk-ai-table --url "<Streamable_HTTP_URL>"

方式二:使用环境变量

export DINGTALK_MCP_URL="<Streamable_HTTP_URL>"

这个 URL 带访问令牌,等同密码,不要泄露。

工作区沙箱

脚本读取本地文件时,会优先使用 OPENCLAW_WORKSPACE 作为允许根目录:

export OPENCLAW_WORKSPACE="$HOME/.openclaw/workspace"

未设置时默认使用当前工作目录。

核心工具集

Base 层

  • list_bases
  • search_bases
  • get_base
  • create_base
  • update_base
  • delete_base
  • search_templates

Table 层

  • get_tables
  • create_table
  • update_table
  • delete_table

Field 层

  • get_fields
  • create_fields
  • update_field
  • delete_field

Record 层

  • query_records
  • create_records
  • update_records
  • delete_records

附件层

  • prepare_attachment_upload

推荐工作流

1. 先找 Base

mcporter call dingtalk-ai-table list_bases limit=10
mcporter call dingtalk-ai-table search_bases query="销售"

2. 再拿 Table 目录

mcporter call dingtalk-ai-table get_base baseId="base_xxx"

3. 再展开表结构

mcporter call dingtalk-ai-table get_tables \
  --args '{"baseId":"base_xxx","tableIds":["tbl_xxx"]}'

4. 字段复杂时读完整配置

mcporter call dingtalk-ai-table get_fields \
  --args '{"baseId":"base_xxx","tableId":"tbl_xxx","fieldIds":["fld_xxx"]}'

5. 再查 / 写记录

mcporter call dingtalk-ai-table query_records \
  --args '{"baseId":"base_xxx","tableId":"tbl_xxx","limit":20}'

mcporter call dingtalk-ai-table create_records \
  --args '{"baseId":"base_xxx","tableId":"tbl_xxx","records":[{"cells":{"fld_name":"张三"}}]}'

6. 写入附件字段

attachment 字段支持三种写法:

方式一:先上传,再写 fileToken(推荐,可靠)

# Step 1:申请上传地址(返回 uploadUrl 和 fileToken)
mcporter call dingtalk-ai-table prepare_attachment_upload \
  --args '{"baseId":"base_xxx","fileName":"report.pdf","size":102400,"mimeType":"application/pdf"}'

# Step 2:把文件 PUT 到 uploadUrl(必须带 Content-Type,值必须与 mimeType 完全一致)
curl -X PUT "<uploadUrl>" \
  -H "Content-Type: application/pdf" \
  --data-binary @report.pdf

# Step 3:把 fileToken 写入记录
mcporter call dingtalk-ai-table create_records \
  --args '{"baseId":"base_xxx","tableId":"tbl_xxx","records":[{"cells":{"fld_attach":[{"fileToken":"ft_xxx"}]}}]}'

方式二:直接传外链 URL(异步转存,best-effort)

mcporter call dingtalk-ai-table create_records \
  --args '{"baseId":"base_xxx","tableId":"tbl_xxx","records":[{"cells":{"fld_attach":[{"url":"https://example.com/file.pdf"}]}}]}'

URL 转存是 best-effort 异步链路,返回成功仅表示已受理,不保证立即可读。可靠写入请用 fileToken 方式。

方式三:原样回传已有附件数据(保留 / 追加已有附件时使用)

query_records 读出的 attachment 单元格数据是完整对象数组,字段形状如下:

[
  {
    "filename": "a.xlsx",
    "size": 92250,
    "type": "xls",
    "resourceId": "<id>",
    "resourceUrl": "<resourceUrl>"
  }
]

其中 type 是文件类别枚举,常见值为 "xls""image" 等;resourceUrl 通常为有时效的下载链接。

如需保留已有附件,把读出的值原样塞回即可。如需追加新附件,把新的 {"fileToken":"ft_xxx"} 与已有对象合并成一个数组一起传入。

update_records 的 attachment 字段格式相同,传入后会整体覆盖该字段。

脚本

批量新增字段

python3 scripts/bulk_add_fields.py <baseId> <tableId> fields.json

fields.json 示例:

[
  {"fieldName":"任务名","type":"text"},
  {"fieldName":"优先级","type":"singleSelect","config":{"options":[{"name":"高"},{"name":"中"},{"name":"低"}]}}
]

兼容项:

  • name 会自动映射为 fieldName
  • phone 会自动映射为 telephone

批量导入记录

python3 scripts/import_records.py <baseId> <tableId> data.csv
python3 scripts/import_records.py <baseId> <tableId> data.json 50

说明:

  • CSV 表头默认按 fieldId 解释
  • JSON 支持:
    • [{"cells": {...}}]
    • [{"fld_xxx": "value"}]

安全规则

  • 文件路径受 OPENCLAW_WORKSPACE 沙箱限制
  • 仅允许读取工作区内 .json / .csv 文件
  • Base / Table / Field / Record ID 都做格式校验
  • 批量上限按 MCP server 实际限制控制:
    • create_fields:最多 15
    • get_tables / get_fields:最多 10
    • create_records / update_records / delete_records:最多 100

调试原则

  • get_base,再 get_tables,必要时 get_fields
  • 不要猜 fieldId
  • 复杂参数一律用 --args JSON
  • singleSelect / multipleSelect 过滤时必须传 option ID,不是 option name

参考

  • API 参考:references/api-reference.md
  • 错误排查:references/error-codes.md

Comments

Loading comments...