What it does
Tiledesk describes itself as an agentic AI operating system. A visual flow builder lets you design AI agents without code, using prompts, prompt chains and MCP tools, and several specialised agents can hand a conversation to one another. Agents draw on knowledge bases that combine full-text and semantic search, with options for multiple knowledge bases, search fallbacks and guardrails. Agents can be published on a web widget, WhatsApp, voice, Facebook Messenger, SMS and email. The site centres on customer support, with escalation to human agents along with full context, a copilot that suggests answers to those agents, and a self-learning loop that adds knowledge extracted from conversations to the knowledge base. Other use cases shown include appointment booking, email ticketing, lead qualification and a Shopify store assistant. It is open source, and the site says it can be deployed on-premises with local open models and a local vector store such as Qdrant, or used as a cloud service. REST and MQTT APIs let you query agents from your own code, and enterprise options include single sign-on.
Capabilities
- No-code visual designer for AI agents using prompts and MCP tools
- Knowledge bases with hybrid full-text and semantic search
- Publishing to web, WhatsApp, voice, Messenger, SMS and email
- Human handoff and a copilot for human agents
- Self-learning from conversations
- On-premises deployment with local models
When to use it
- You want to build customer support agents without writing code
- You need one agent across several channels such as WhatsApp and web chat
- You must keep data on your own infrastructure and use open models
