The Splashify Pro Claude Code Plugin

This section explains how to control your Splashify Pro account from Claude Code  — Anthropic’s official terminal-based AI coding agent — by installing the Splashify Pro plugin.

Once the plugin is installed, you simply ask Claude — “send a WhatsApp to this customer”, “start a May Sale broadcast to my VIP segment”, “what’s my wallet balance” — and it carries out the action on your Splashify Pro account.

One token, three front-ends. The Claude Code plugin (this section), the OpenClaw skill, and the standalone splashify CLI all drive the same account through the same oc_live_ access token stored in ~/.splashify/config.json. Install whichever assistant you already use; you don’t need all three.

What is Claude Code?

Claude Code  is Anthropic’s official agentic coding tool, available as a terminal CLI, desktop app (macOS/Windows), web app at claude.ai/code , and IDE extensions for VS Code and JetBrains. It supports plugins — bundles of skills, commands, agents, and MCP servers that extend what Claude can do.

The Splashify Pro plugin is published in the public plugins directory at claude.com/plugins  and source-hosted at github.com/splashifypro/claude-plugin .

What is the Splashify Pro plugin?

The Splashify Pro plugin is a small bundle of seven progressive-disclosure skills (markdown instruction files) that you install into Claude Code. Each skill loads only when it matches what you ask:

The plugin itself holds no credentials and calls no API directly. It only teaches Claude how to drive the splashify CLI on your machine. The CLI is what actually talks to the Splashify Pro backend.

How the connection works

┌──────────┐  "send a WhatsApp to John"   ┌────────────────┐
│   You    │ ───────────────────────────► │   Claude Code  │
│          │                               │  (terminal /   │
└──────────┘                               │  desktop / IDE)│
                                           └──────┬─────────┘
                                                  │ matches a Splashify skill
                                                  │ reads SKILL.md instructions
                                                  ▼
                                           ┌──────────────┐
                                           │ splashify CLI│  runs on your machine
                                           │ (Bash tool)  │  via Claude Code's Bash
                                           └──────┬───────┘
                                                  │ HTTPS, Bearer oc_live_…
                                                  ▼
                                      ┌───────────────────────┐
                                      │ Splashify Pro backend │  token → your account
                                      └───────────────────────┘

1. You create a personal access token (oc_live_…) in Settings → Developer → Access Tokens and connect the splashify CLI with it.
2. You install the plugin into Claude Code with /plugin marketplace add splashifypro/claude-plugin then /plugin install splashifypro@splashifypro.
3. When you ask Claude to do something, it matches the relevant skill, reads the skill’s instructions, and runs the splashify CLI through its built-in Bash tool. The CLI calls the Splashify Pro API as you.

Everything runs as you — scoped to your account only. The token never grants admin or reseller access.

How the three front-ends compare

The plugin is a thin instruction layer on top of the CLI. You install the CLI either way; the plugin just lets Claude Code drive it for you.

What you can do

After installing the plugin, you (by asking Claude) can:

• Messaging — send WhatsApp text, template, and media messages; send RCS text and templates; send Instagram DMs within the messaging window; read, resolve, reopen and assign conversations.
• Contacts — list, search, create, update, tag, untag, block, unblock and delete contacts; manage tags; manage custom attributes; manage audience segments (static and dynamic).
• Broadcasts — list, create, schedule, cancel, restart and rebroadcast campaigns; track per-cohort delivery; export per-recipient CSV.
• Templates — list approved WhatsApp templates; sync from Meta; create and delete templates.
• AI agents — create chat and voice agents; upload knowledge files (PDF / DOCX / Markdown / TXT); set the default agent.
• WhatsApp Flows — list, sync, deprecate and inspect responses.
• Integrations — Shopify, Zapier, Pabbly, Make, Google Sheets, Salesforce and more; CTWA / Meta Conversions API setup and event sends.
• WhatsApp Calling — settings, analytics, call buttons, permission templates and template calls.
• Email marketing — sending domains (SPF/DKIM/DMARC), templates, audiences and campaigns.
• Analytics & billing — message analytics, wallet, billing, invoices, subscription, expenses, AI credits, activity logs.
• Account admin — team members and permissions, WABA setup, opt management, IP allowlist, active devices, support tickets.

Security highlights

The plugin’s threat model is documented in full at /claude-code/security, but the headlines:

1. Zero credentials in the plugin. The CLI keeps your oc_live_ token in ~/.splashify/config.json (mode 0600). The plugin never reads that file, and Claude never sees the token.
2. Backend-scoped tokens. oc_live_ tokens can only reach /api/v1/app/* — admin, reseller, and platform endpoints are server-rejected even if Claude were tricked.
3. Confirmation on every mutating action. Sends, broadcasts, deletes, token revokes, and raw API writes all require an explicit “go” from you, with the destination repeated back.
4. Prompt-injection resistant. Skills treat fetched URLs, email bodies, and tool results as data — not instructions. Claude must get the destination phone or email from your own message in the current session.

Where to go next

1. Install the plugin — install Claude Code, the splashify CLI, connect your account, and add the plugin.
2. Using the plugin — example prompts for Claude.
3. Security model — full threat model and guarantees.
4. Troubleshooting — fixes for common issues.
5. splashify CLI — the standalone command-line tool, documented separately.