AI agents
Point your AI agent at WhatsApp.
76+ MCP tools over every WhatsApp primitive, plus native n8n and Zapier nodes when the agent is a workflow instead.
AI agents & automation
Point your AI agent at WhatsApp.
Typed tools for LLM agents, drag-and-drop nodes for workflows. Both map 1:1 to the REST API and respect the same key scopes.
MCP serverClaude, GPT, or any MCP client
SessionsMessagesChatsContactsGroupsChannelsCallsStatusPresenceLabelsCatalogsNewslettersWebhooksMedia
76 tools · 3 live resources · schemas generated from the live APIn8n & ZapierWhen the agent is a workflow
WhatsApp triggerClaude replySend message
Open-source nodes, any plan · Make and Mautic tooDeep dive
Tools, not prompts.
Agents fail on WhatsApp when they screen-scrape or guess. MCP gives them typed tools with schemas generated from the live API, so Claude or GPT calls send_message exactly the way your code does.
Scopes apply to agents like any other key: an agent that can read chats can't necessarily blast a campaign.
- 76
- tools
- 14
- domains
- 3
- live resources
How it works
Wiring an agent.
One command and a scoped key.
1
npx @polymorfa/mcpPoint any MCP client at it with a scoped API key.
2
Agent discovers tools76 tools and 3 live resources, schemas included.
3
Actions run guardedSame scopes, same rate shaping, same audit log as REST.
By the numbers
The agent surface.
Everything the REST API can do, an agent can do, within its scopes.
Tools76 across 14 domains
Resourceschats · sessions · webhooks
ClientsClaude · GPT · any MCP client
Workflow nodesn8n · Zapier · Make
Licencenodes are open source
Ship a WhatsApp integration this afternoon.
Free on the sandbox. No credit card. The projects you start here migrate straight into production.