WeCom (Tencent's enterprise messaging platform) has added CLI and MCP support to its open capabilities in version 5.0.10, giving AI agents a more product-like entry point into office software integration. According to the official announcement, the newly opened capabilities cover 10 core office modules, and WorkBuddy, DeepSeek Harness, and enterprises' self-built agents can all connect within the scope of their authorization.
The move brings AI out of the chat window and into the call chain of business tools. The CLI is suited to embedding actions into scripts and workflows, while MCP gives different models and agents a relatively unified way to make calls. Tasks like document editing, spreadsheets, data analysis, proposal writing, and business dashboards can move from "copying a pasted answer" to "completing a step directly inside the business system."
Before opening up, permissions had to be split apart first
In its announcement, WeCom listed four safeguards: isolating human and AI permissions, requiring manual approval for key operations, time-limited authorization, and full audit trails. These four constraints deserve closer scrutiny than simply asking "how many agents does it support" — reading a spreadsheet and bulk-editing a batch of customer records don't carry the same level of risk, and letting a model call a tool is not the same as handing it an employee account's full set of permissions.
For enterprise IT teams, the next step is verifying whether approval workflows can cover high-risk actions, whether audit logs can reconstruct both parameters and results, and whether administrators can set different tool boundaries for different departments. Once the interface is open, agents will move into office systems faster — and if permission design doesn't keep pace, mistakes get amplified by automation too.
Sources: WeChat Open Class, CocoLoop; verified against WeCom 5.0.10's CLI, MCP, 10 office modules, and authorization/audit mechanism.