Coding Agent
Reading guide
Parent section: Channel management. Search keywords: Coding Agent, Codex, Claude Code, Grok Build, OpenCode, WorkBuddy, 豆包, 千问, CLI, Skill.
- Full installation, commands, and usage scenarios: CLI user guide.
- Learn how chat platform channels are connected and bound: Channel overview and binding.
Besides using Agentic Engine in chat platforms such as Feishu and Slack, you can also use it in coding agents and desktop AI assistants on your own computer, such as Codex, Claude Code, Grok Build, OpenCode, WorkBuddy, 豆包工作, and 千问办公. These agents connect to your Agentic Engine environment through Agentic Engine CLI (ae-cli) and, within the scope of your account permissions, query data, build dashboards, manage tracking, and follow up on engagement and data development tasks.
How it differs from chat channels: For channels such as Feishu and Slack, an admin creates a channel instance in System Admin, and then users bind their accounts. Coding agents don't need an instance created in channel management. Each user installs the CLI on their own computer and authorizes it with their own account.
Requirements
- The agent can run terminal commands on your computer: the agent calls the CLI in your local terminal.
- Node.js 20 or later is installed on your machine.
- You have the Agentic Engine address, and your account has access to the target project.
Supported agents
The companion Skills are installed by the open-source skills installer. It automatically detects which coding agents are installed on your machine and installs the Skills into each agent's directory. The following table lists common agents' names in the installer and their global Skills directories:
| Agent | Name in the installer | Global Skills directory |
|---|---|---|
| Codex | codex | ~/.codex/skills/ |
| Claude Code | claude-code | ~/.claude/skills/ |
| Grok Build | grok | ~/.grok/skills/ |
| OpenCode | opencode | ~/.config/opencode/skills/ |
| Cursor | cursor | ~/.cursor/skills/ |
| Trae、Trae CN | trae、trae-cn | ~/.trae/skills/、~/.trae-cn/skills/ |
To install for only some of these agents, specify them with -a:
npx -y skills add ThinkingAIAgenticEngine/ae-cli -g -y -a codex -a claude-code
Desktop agents such as WorkBuddy, 豆包工作, and 千问办公 aren't on the installer's list. As long as they can run terminal commands, you can install the CLI by following the steps below. If they can't load Skills automatically, tell the agent at the start of the task: "Use ae-cli to do this. First run ae-cli --help to see the available commands."
Setup steps
Install the CLI and Skills
In the agent you want to use, send the following text, and replace the address in it with your Agentic Engine address:
Read https://raw.githubusercontent.com/ThinkingAIAgenticEngine/ae-cli/main/cli-installation-guide.md and follow its instructions to install or upgrade Agentic Engine CLI (ae-cli) for me. My Agentic Engine address is https://your-ae-host.example.com. Install only ae-cli and the companion Skills, and don't modify the current project. If you need me to log in and authorize, or if admin permissions are required, pause and let me know.
You can also install them manually in the terminal:
npm install -g @thinkingai/ae-cli
npx -y skills add ThinkingAIAgenticEngine/ae-cli -g -y
Log in and authorize
The agent sends you the authorization link. If you installed manually, run:
ae-cli auth login --host https://your-ae-host.example.com
ae-cli auth status --host https://your-ae-host.example.com
In your browser, check the verification code and confirm the authorization. When authenticated is true in the login status, you're done. Credentials are stored on your machine by Agentic Engine address. When you switch to another agent on the same computer, check the login status first, and authorize only if you aren't logged in.
Reload and verify
After installation, close and reopen the agent, or start a new task, so that it loads the Skills. Then send:
Check whether Agentic Engine CLI is available: confirm the current version, login status, and Agentic Engine address; then list the first 5 projects I can access, returning only the project name and project ID. Only view for now; don't modify anything.
If the project list is returned, setup is complete. Next, you can ask questions based on the scenarios in the CLI user guide.
Usage tips
- In each task, state the project, the objects, the time range, and how far the agent is allowed to go, for example, "Read-only, don't modify anything" or "Give me a plan first, and run it only after I confirm".
- Have the agent include the events, definitions, time range, and filters it used in its conclusions so that you can verify them.
- Everyone on the team authorizes with their own account; don't share tokens.
- For tasks that need to run regularly, run them manually once first, and then use the agent's built-in scheduling feature or Automations in Agentic Engine.
FAQ
What if the agent treats my question as ordinary chat?
Usually the Skills haven't been loaded. Confirm that the Skills are installed, and then restart the agent. If that still doesn't work, say "Use ae-cli to do this" directly in your question.
What if I get "command not found" or a version mismatch?
Run ae-cli update --host <address> to sync the CLI and Skills versions to what the environment requires, and then restart the agent.
What if my company computer doesn't allow global npm package installation?
Use nvm, fnm, or Volta to install Node.js under your user directory, and don't use sudo npm install -g. If company policy still blocks it, contact your IT admin.
What's the difference between coding agents and Agents in the platform?
Agents in the platform are used in Agentic Engine workspaces, conversations, and chat channels; for configuration, see Agent. Coding agents run on your own computer, access Agentic Engine through the CLI, and can be combined with local files, code repositories, and other tools (such as lark-cli).