Put WhatsApp AI agents to work in sales and support.

What an agent really is

An AI agent combines a model, instructions, tools, and context.

The model understands the conversation. Instructions set the boundaries. Tools let the agent do approved work, and context helps it continue a task across several messages. Waaru brings these parts together so your team can manage them in one place.

Tool calling

Give the agent a clear set of approved actions.

Each action specifies the information it needs, such as an order reference. Waaru checks that information before calling the connected system and records the result for review. Invalid input returns an error for the agent to handle.

• WhatsApp: send a message, template, or flow and list conversations
• Custom REST: define an endpoint with an OpenAPI specification
• Handoff: send a reason, priority, and summary to your team
• Business integrations: check each connector page for its supported actions and setup requirements
• Planned integrations: HubSpot, Pipedrive, Freshdesk, and Zendesk

MCP-native

Connect an existing AI agent through Model Context Protocol.

Model Context Protocol, or MCP, gives AI applications a standard way to discover and use tools. Waaru exposes approved WhatsApp operations as documented, workspace-scoped tools. MCP does not set permissions or approval rules by itself, so you still decide which tools an agent can use and what your team must review.

Memory

Keep the right context for the customer and your business.

Conversation context covers the current exchange. Contact context can retain approved details that help in a later conversation, such as a language preference. Workspace context covers shared business information such as policies and frequently asked questions. Keeping these scopes separate makes the information easier to manage and review.

Governance

Set the actions the agent can take on its own.

Workspace rules define which tools the agent may call, which actions need human approval, what it must refuse, and when it should escalate. These controls let your team give the agent useful work without giving it open-ended access.

• Tool allowlists limit each workflow to the tools it needs
• Approval rules pause sensitive actions for human confirmation
• Refusal rules define topics or actions the agent must decline
• Writing rules set tone, required wording, and restricted phrases
• Rate limits cap tool calls by contact and workspace

Handoff

Give your team the context to take over.

Every agent has a confidence threshold and an explicit escalate tool. When the agent escalates, or when a customer asks for a person, the conversation is routed to the team inbox with the transcript, an AI summary, the contact record, and prior notes. After the teammate resolves the case, the approved resolution can become workspace-scoped context for a similar future intent.

Observability

Review what the agent said and did.

Each conversation produces a structured trace. It records the model, the tools considered, tool inputs and results, the reply, the confidence score, and the policy decisions involved. Your team can inspect the trace in the dashboard or through the API, then adjust a tool or rule when a conversation goes wrong.