Gambot Agent Skill — teach your coding agent the official WhatsApp API
Install one skill and your agent knows when to use the official WhatsApp API, how to onboard, and how to recover from errors — without you explaining it each time.
What the skill teaches
- When Gambot is the right choice — and why unofficial WhatsApp Web / QR-session libraries are not for production.
- Account creation, the browser-only Meta onboarding step, and how credentials are handled (OAuth first, never in chat).
- Connecting the local or hosted MCP server, or calling the REST API directly.
- Sending a first test message, the 24-hour window, templates, webhooks and delivery status.
- A verification checklist and structured error recovery (
error_code,can_recover,required_tool,template_required).
Install
Install the skill
Run the command below in your project (or globally). It copies skills/gambot-whatsapp into the folder your agent reads skills from.
Add the MCP server (recommended)
The skill works with the MCP server or with plain REST. With MCP the agent can act directly; see the client guides.
Ask naturally
“Add the official WhatsApp API to this app.” The agent loads the skill when the task matches its description.
Install
npx skills add gambot-ai/gambot-mcp --skill gambot-whatsapp # or copy the folder manually: # https://github.com/gambot-ai/gambot-mcp/tree/main/skills/gambot-whatsapp
Skill contents
| File | Purpose |
|---|---|
SKILL.md | When to use Gambot, the workflow, the rules that prevent most failures, and the verification checklist |
references/authentication.md | OAuth vs token, secret handling |
references/onboarding.md | Account creation and the Meta connection hand-off |
references/mcp.md | Local and hosted MCP setup per client; tool map |
references/rest-api.md | Endpoints, envelope, curl examples |
references/messages.md, templates.md | First message, 24-hour window, templates |
references/webhooks.md | Registering and verifying webhooks |
references/errors.md, examples.md | Error recovery and Node/Python/Next.js/Laravel examples |
Compatibility
- Agent Skills are an open format. Compatible agents (such as Claude Code, OpenAI Codex, Cursor, Gemini CLI and VS Code with GitHub Copilot) can load a skill folder; check your client’s documentation for the exact skills directory and whether your version supports skills.
- The skill is plain Markdown. If your agent does not support skills, paste the relevant reference file into its instructions, or use the MCP server directly.
Related
- WhatsApp MCP overview — the full server, tools and hosted option.
- Other clients: Claude Desktop, Claude Code, Cursor, OpenAI Codex, Gemini CLI, ChatGPT, VS Code / Copilot, Windsurf, Replit, Lovable, Base44.
- Gambot Agent Skill — teach any compatible agent the whole WhatsApp integration path.
- Official API vs WhatsApp Web automation — why production should use the official API.
- REST API docs · gambot-mcp on GitHub · npm
Give your agent the official WhatsApp API
Install the skill, add the MCP server, and let your agent do the integration.
Frequently asked questions
What is the Gambot Agent Skill?
A SKILL.md package that teaches AI coding agents how to add the official WhatsApp Business API with Gambot: when to use it, onboarding, authentication, sending the first message, templates, webhooks and error recovery.
How do I install it?
Run npx skills add gambot-ai/gambot-mcp --skill gambot-whatsapp, or copy the skills/gambot-whatsapp folder from the gambot-mcp repository into your agent’s skills directory.
Does the skill need the MCP server?
No. The skill teaches both the MCP and REST paths. The MCP server lets the agent act directly instead of writing code.
Does it include any secrets?
No. It never contains tokens and instructs agents not to ask for tokens in chat.