FAQ

How do I add my agent to my website?

Go to Settings → Deploy and copy the Widget Embed Code. Paste it before the </body> tag on your website. Works on any platform — Webflow, WordPress, Shopify, or custom sites.

What file types can I upload to a knowledge base?

PDF, DOCX, CSV, and Excel files. You can also scrape websites or paste text manually.

Can users send files in the chat?

Yes. Users can attach PDFs, images (JPG, PNG, GIF, WebP), CSV files, and text or Markdown files. The AI reads the content and responds based on what's in the file.

Can I use my own AI model?

In Agent Settings → AI, you can choose from OpenAI (GPT-4o, GPT-4o mini) and Anthropic (Claude). Support for custom LLM endpoints is on the roadmap.

Can I have multiple knowledge bases per agent?

Yes. An agent can use any number of knowledge bases and searches across all of them.

What happens when I unpublish an agent?

The landing page and widget stop working immediately. Your configuration is preserved and you can republish at any time.

Can I use a custom domain?

Custom domain support (e.g. chat.yourbrand.com) is coming soon for Scale+ plans.

How does the Setup Assistant work?

When you sign up, the Setup Assistant guides you through creating your first agent via conversation. You can also access it any time via the Build with AI button inside the agent builder.

What's the difference between Knowledge and Action agents?

Knowledge agents are conversational — they chat with users in real time via a chat widget or landing page, drawing from your knowledge bases and journeys. Action agents are autonomous — they run workflows in the background on triggers (schedules, webhooks, data changes) without user interaction. Use Knowledge agents for support, recommendations, and lead capture. Use Action agents for automation, monitoring, and document generation.

Do action agent runs cost credits?

Only nodes that use LLM processing consume credits (AI Summarize, Classify, Extract Data, Generate Document — 1 credit each per execution). Non-AI nodes like Send Email, Fire Webhook, Delay, and Condition are free. Test and dry runs are always free.

Can I test action agents before publishing?

Yes. Go to the Overview tab and click Test Run. You can provide a mock trigger payload and view the full execution trace showing input/output data and timing for each node. Test runs use draft configuration and don't consume credits.

How do I connect WhatsApp?

Go to Connections, find WhatsApp, and click Connect. Enter your 360dialog API key. Zimmer validates the key and automatically registers the webhook. Then link your WhatsApp integration to an agent from the agent detail page.

How do I sync Notion content?

Go to Connections, find Notion, and click Connect. Complete the OAuth flow to authorize Zimmer. Once connected, you can select specific pages and databases to sync, or sync everything. Content is automatically chunked and embedded for semantic search.

What integrations are available?

Currently available: Google Drive, Instagram, Facebook Page, Notion (knowledge sources), Google Sheets, Shopify, Custom API (data sources), WhatsApp (channel), Google Calendar (booking), GA4, Webhooks, Slack, Mailchimp (analytics and events). More integrations are coming soon.

What are journey kits?

Journey kits are pre-built conversation flow templates you can apply to any agent. They include complete node structures with AI categorizers, quick replies, data collection, and escalation paths. Available from the Library in the left nav. Categories include e-commerce, support, lead generation, enterprise, and agency.

What's the difference between a Starter Kit and a journey from the Library?

A Starter Kit (from the Kits page) creates a complete agent with template design, AI config, system prompt, and a journey flow all bundled together. A Library journey is just the conversation flow — you apply it to an existing agent. Use starter kits to get started quickly; use library journeys to swap flows on existing agents.

How do I control who can access which projects?

Go to Settings → Members. Owners and Admins see all projects automatically. For Editors and Viewers, click the project access column to assign them to specific projects. When inviting new members, you can select which projects they get access to.

What do journey diagnostics do?

The journey builder runs continuous validation on your flow and shows errors and warnings with fix hints. Errors (like dead-end nodes or disconnected nodes) must be fixed before publishing. Warnings (like missing error handling or unused condition branches) are recommended fixes. Each diagnostic includes a plain-language hint explaining exactly how to resolve the issue.

How do conversation review queues work?

The Conversations page groups conversations that need attention into queues: abandoned, negative sentiment, leads captured, handover requests, knowledge gaps, and reusable answers. You can assign conversations to team members, add internal notes, and flag issues for follow-up.

Can agents talk to each other?

Yes. Zimmer supports multi-agent coordination in two ways. Agent Groups let agents share context passively — a Knowledge agent in the same group as an Action agent can see its recent run results. The Trigger Agent node lets one agent invoke another directly from a journey flow, either waiting for the result (sync) or firing and forgetting (async).

How do I set up Agent Groups?

Go to any agent's Settings → Groups. You can create a new group, join an existing group, or leave a group. All agents in the same group automatically share context. For example, add a "Sales Team" group containing your Lead Enricher action agent and your Sales Assistant knowledge agent — the knowledge agent will be aware of enrichment results when answering questions.

What is the Trigger Agent node?

The Trigger Agent node is a flow node that invokes another agent's headless run from within a journey. You configure the target agent, map input variables, choose sync or async mode, and set a timeout. It's available under the Cross-Agent category in the node palette. Each triggered agent consumes its own credits. Trigger chains are limited to 3 levels deep to prevent infinite loops.

Where can I get help?

Use the ? help button in the bottom-right corner of the dashboard to ask questions about Zimmer. For support, email hello@zimmer.ai.