Base de conocimiento de GetMyBot

Skills for AI Agents

Ready-made GetMyBot skills for Claude Code and other agents: manage bots, reactions, and broadcasts with natural language.

En esta página

GetMyBot skills are ready-made agent wrappers around the GetMyBot API. They teach an AI agent (for example, Claude Code) to work with your bots in natural language: "create a reaction in GetMyBot", "launch a GetMyBot broadcast to the VIP label", "show bot stats". Under the hood, a skill calls GetMyBot MCP tools, and if the MCP server is not connected it makes the same requests over REST with your token.

Skills are not a separate paid feature and not a marketplace. They are SKILL.md files (Claude Code format) that you install locally next to your agent.

How it works

Each skill is built on the principle of progressive disclosure: a short description triggers the skill on domain phrasing, while the heavy API reference is loaded only when needed. Inside each skill, two access paths are documented:

  • MCP (preferred): if the GetMyBot MCP server is connected to the agent, the skill calls tools directly.
  • REST fallback: if MCP is unavailable, the skill performs the same operations via curl requests with a personal token.

Authentication across all skills is the same: a personal mbp_… token and the base address /api. See the authorization guide for details.

Six skills

  • mybot: entry point and navigation: what the GetMyBot API is, how to get a PAT and its scopes, the base address, how to connect the MCP server, when to use REST vs. MCP, error format, pagination, and limits. From here the agent navigates to the other five skills.
  • mybot-bots: connecting and managing bots: token verification, connection, list and card, update, sync, token change, transfer, deletion, settings, templates.
  • mybot-reactions: the reaction builder (the largest skill): "trigger → conditions → chats → actions" structure, creation and update, trigger types, AND/OR conditions, action catalog, messages and buttons, dynamic buttons and parameters, formulas (DSL), folders, reaction links, import from SamBot, formula validation, ordering.
  • mybot-broadcasts: broadcasts: launching by reaction, segmentation by labels, pause/resume/cancel and status, editing already-sent messages (single and "for everyone"), queue and retries.
  • mybot-data: data and integrations: collections (tables: fields and records), scenarios (flows), integrations (AI, Google Sheets, GetCourse, external handler via WebSocket), connections and access.
  • mybot-analytics: observability and audience: stats (balance spend, metrics, log, reaction health), analytics (audience, chats, funnels), subscribers and labels, dialogs, billing (balance, plans, transactions).

Installation

Skills live in the GetMyBot repository inside the skills/ folder. Each skill is a skills/<name>/ directory with a SKILL.md file (quick guide) and a reference.md file (full domain reference with MCP and curl examples).

To make the skills visible to your agent, link them to your Claude Code skills directory. The repository includes a ready-made script:

./skills/install.sh

It creates symlinks ~/.claude/skills/<name> pointing to the folders inside the repository: edits in the repo are immediately visible to the agent, with no extra copies. After installation, state your task in plain language ("build an auto-reply bot in GetMyBot") and the agent will pull in the right skill automatically.

New customer workflow skills

  • mybot-ai: Managed AI invoke/usage and the safe Copilot propose → validate → simulate → apply lifecycle.
  • mybot-intelligence: Flow Intelligence policy, bounded reaggregation, Explainable Replay, and regression fixtures with non-causal interpretation rules.
  • mybot-playbooks: Vertical Playbooks, resumable setup, preflight, atomic install, safe upgrade, KPI contracts, and outcome onboarding.
  • mybot-channels now includes Facebook Messenger and Instagram Direct OAuth handoff, account discovery, checks, selection, attachment, and pending-connection editing.

These skills expose only customer-safe operations. They cannot access Managed AI provider keys, routing, prices, or reconciliation.

What's next