Qyvo

Qyvo exposes its WhatsApp marketing platform as a Model Context Protocol server. Connect any MCP-compatible client and an LLM agent can list and tag contacts, draft and schedule broadcasts, trigger sequences, reply in the inbox, and read analytics — all bounded by your workspace.

Endpoint

https://www.qyvo.io/mcp

Aspect	Value
Transport	HTTP (Laravel `laravel/mcp` package, MCP 2024-11 spec)
Auth	`Authorization: Bearer <token>`
Server name	`Qyvo`
Server version	`1.0.0`
Tools	35, grouped by resource
Prompts	4 (Campaign Assistant, Segmentation, Performance, Inbox Triage)
Resources	1 (Meta WhatsApp guidelines)

All operations are scoped automatically to the workspace that owns the token. You cannot read, write, or send anything outside it.

Why MCP, not the REST API?

The REST API and the MCP server expose the same workspace — same tenant, same data, same write semantics. Two reasons to pick MCP over REST:

Tool descriptions designed for LLMs. Each tool ships with a description tuned to help the agent pick the right one. The annotation IsDestructive(false) lets clients (Claude Code, Cursor) auto-approve safe operations without prompting on every call.
Resources and prompts. The MCP server publishes a Guidelines resource that reminds the agent of Meta's policy (24h window, opt-in, template categories), and four Prompts that pre-bake common workflows.

If you're writing your own application code, REST is simpler. If you're connecting an LLM agent, MCP is the right surface.

Quick paths

Quickstart — generate a token and verify the handshake.
Connect a client:
- Claude Desktop
- Cursor
- ChatGPT
- n8n
Tools reference — the full 35-tool catalogue, by resource.
Example prompts — what to ask the agent once connected.

Scope and safety

Every tool resolves the tenant from the bearer token. There is no way to operate cross-tenant.
Read-only tools (list_*, get_*) are annotated IsReadOnly and IsIdempotent. Clients that respect annotations auto-approve them silently.
Side-effectful but reversible tools (create_*, update_*, tag_*, trigger_*, send_message, import_contacts, sync_templates) are annotated IsDestructive(false). Most clients still prompt the first time, then offer "always allow".
Truly destructive tools (delete_contact, delete_segment, delete_broadcast, delete_template) are annotated IsDestructive and clients prompt for confirmation on every call.