Recipes

Recipes are prebuilt workflow templates. Instead of assembling triggers, agents, and tools on an empty canvas, you pick a recipe that matches what you want, click Create, and get a complete working workflow you can then customize. Every recipe is a normal workflow once installed; nothing about it is locked.

If you have never built a workflow before, start here. Installing a recipe and reading its diagram is the fastest way to understand how workflows fit together.

Installing a recipe

1. Open Workflow in the navigation. The page lands on the Workflow recipes tab, a gallery of cards. Each card shows the recipe's name, description, a difficulty badge (beginner or intermediate), and an estimated setup time. Use the search box or the category filters to narrow the list, and the small gear icon on a card to read its full details.
2. Click Create on the card you want. Faberiq creates any collections the recipe needs, creates the workflow from the template, and opens it in the designer.
3. If the recipe needs connectors or a knowledge base, the setup wizard opens automatically. It lists each requirement as a step: pick an existing connector or connect a new one, and optionally upload a starter document for the knowledge base. You can skip steps and finish them later; the designer reminds you when a workflow still has unbound connectors.
4. Save the workflow and test it from the chat, or from the channel you connected.

Prefer to start from nothing? The Blank workflow button next to the search box creates an empty workflow and opens the wizard.

Starter bundles

A Starter bundles section sits at the top of the recipe gallery. A bundle installs several workflows and their supporting pieces in one click, so a new account gets a complete setup instead of assembling it recipe by recipe.

Faberiq starter bundle

One install gives you a working front office:

• The AI Front Desk workflow (the same one as the recipe below).
• The Follow-up Agent workflow, daily schedule included. See Follow-up Agent.
• A Customer journey · Signup paperwork onboarding workflow that runs when a customer joins the portal.
• Three published tasks: Customer intake, Non-disclosure agreement, and Terms & conditions. The NDA and Terms & Conditions ship as editable markdown templates with [Company Name] placeholders; edit them under Workflow → Tasks → Task settings before customers sign them.

Installing also wires the organization defaults on the Setup tab — the default front desk workflow, the default follow-up workflow, and the onboarding workflow with its task toggles — wherever you have not already chosen a value. Re-installing never overwrites your configuration, so the bundle can restore a lost setup at any time.

After install, the bundle surfaces the connectors it needs: Instagram Messenger (the front desk and follow-up channel), Google Calendar (appointments), and Square (quotes and invoices).

The recipe catalog at a glance

Recipe	Difficulty	Setup	Connectors required
AI Front Desk	Intermediate	10 min	Instagram Messenger, Google Calendar, Square
Telegram Bot	Beginner	3 min	Telegram
WhatsApp Business Bot	Beginner	3 min	WhatsApp
SMS Chatbot	Beginner	3 min	Twilio SMS
Slack Bot	Beginner	3 min	Slack
Facebook Messenger Bot	Beginner	3 min	Facebook Messenger
Instagram Messenger Bot	Beginner	3 min	Instagram Messenger
Email Support Agent	Beginner	3 min	Gmail
Knowledge Base Chatbot	Beginner	2 min	None
API Chatbot Endpoint	Beginner	2 min	None
Meeting Vault	Beginner	5 min	None
Google Reviews Responder	Intermediate	5 min	Google Reviews
Sync data source to a collection	Beginner	3 min	Financial Market Prep
Analyze a collection with AI	Intermediate	5 min	None

Setup times are the estimates shown on the cards and assume the required accounts already exist. The rest of this page walks through each recipe: what it does, what it needs, and what to customize.

AI Front Desk

The flagship recipe: one agent that answers questions, converts leads into customers, books appointments, and sends Square quotes and invoices that collect payment.

What it does. Customers reach the front desk on any connected channel; Instagram DMs are wired out of the box. The workflow has:

• Channels as entry points. Each channel is its own trigger state (the installed recipe starts with an Instagram trigger). A trigger fires when a message arrives and hands the conversation to the agent. To add Telegram, Messenger, or WhatsApp, check them under Channels in Workflow Settings; each becomes another entry point feeding the same agent.
• The Front Desk agent and its playbook. The agent's system prompt is a step-by-step playbook: answer questions with knowledge base search; the moment a visitor shows real interest, collect their name and email, check customer.find, and convert them with customer.create; for meetings, confirm date, time, and timezone, then call appointment.book; for pricing, confirm the line items, then call invoice.create (a quote returns a payment link to share in chat, an invoice is emailed by Square); check invoice.status before following up on payment; end meaningful conversations with a knowledge.note so the next conversation has context; and always reply through the respond tool of the channel the conversation came from. The agent runs with a generous budget (16 turns, 24,576 tokens, 3 minutes).
• Success and fail branches. The agent's next transition points at a Succeed node and its catch transition at a Fail node, so a conversation the agent could not handle ends as a visibly failed run with the message "Front desk agent was unable to handle the message."

What it needs. Three connectors: instagram_messenger (the inbound channel), google_calendar (real calendar invites for bookings), and square (quotes, invoices, and payment links). The wizard walks you through all three. A knowledge base with your services and prices makes the answers and quotes accurate.

What to customize. Add more channels; edit the playbook to match your business (deposit rules, service areas, tone); point the fail branch at a step that alerts a human instead of just failing; and fill the knowledge base so the agent quotes real prices.

Follow-up Agent

The front desk's memory for money left on the table. It installs as part of the Faberiq starter bundle.

What it does. Every day at 09:00 a schedule trigger starts the agent, which:

1. Calls followups.list for today's candidates: quiet leads, quotes waiting on an answer, and unpaid invoices, per the cadence set on the Setup tab.
2. Writes a short personal message for each one: a friendly check-in for a lead, a nudge with the quote link for a waiting quote, a polite payment reminder with the payment link for an unpaid invoice.
3. Sends each message with the respond tool for the candidate's channel, using the routing followups.list provides, so an Instagram lead is nudged on Instagram.
4. Records each touch with followups.record, so the cadence window restarts.
5. Finishes with a knowledge.note summarizing how many follow-ups went out and anything that needs a human.

What it needs. The channel connectors your leads arrive on, and follow-ups enabled under Workflow, Setup. Customers whose journey is marked Completed or Lost are never contacted.

What to customize. The cadence (days per stage and the per-run cap) on the Setup tab, the schedule time, and the message tone in the agent's prompt. The tools it runs on are documented in the Agent tool reference.

Channel chatbots

Six recipes share one simple shape: a channel trigger receives the message, a Knowledge Assistant agent answers it, and a reply step sends the answer back on the same channel. The agents search your knowledge base, and most also carry the front desk basics (customer.find, customer.create, appointment.book, knowledge.note), so they can convert an interested visitor mid-chat.

Telegram Bot

What it does. Answers Telegram messages from your knowledge base and can convert leads and book meetings. Needs a telegram connector (a Telegram bot token). Customize the agent's system prompt and the knowledge base it searches.

WhatsApp Business Bot

What it does. Handles customer conversations on WhatsApp with the same knowledge assistant shape. Needs a whatsapp connector. Customize the prompt, tone, and knowledge base.

SMS Chatbot

What it does. Receives text messages and replies over SMS. Needs a twilio_sms connector. Customize the prompt; keep answers short, since this is SMS.

Slack Bot

What it does. Answers questions in Slack channels from your knowledge base. Needs a slack connector with a bot token. Customize the prompt, and optionally give the agent the wider Slack tool family to post to channels or send DMs.

Facebook Messenger Bot

What it does. Automated customer chat on your Facebook Page's Messenger. Needs a facebook_messenger connector with messaging permission. Customize the prompt and knowledge base.

Instagram Messenger Bot

What it does. Automated replies to Instagram DMs. Needs an instagram_messenger connector. Customize the prompt and knowledge base; upgrade to the AI Front Desk recipe when you want bookings and payments too.

Email Support Agent

What it does. Reads incoming email, searches your knowledge base for the answer, composes a helpful reply with the agent, and sends it back as an email reply. The flow is: Gmail trigger, then a knowledge base search step, then the composing agent, then the Gmail reply step.

What it needs. A gmail connector, plus a knowledge base holding your support content.

What to customize. The agent's reply style and sign-off, and the knowledge base folders it searches.

Knowledge Base Chatbot

What it does. The simplest RAG chatbot: a Conversation API trigger receives the question, the agent searches the knowledge base and answers, and an output node returns the response. This is the recipe behind a plain "answer questions from my documents" assistant, usable from studio chat, portal chat, and the web chat iframe.

What it needs. No connectors. Just upload documents to the Knowledge Base.

What to customize. The system prompt (it says to admit when the answer is not in the knowledge base; keep that) and the starter prompts in Workflow Settings.

API Chatbot Endpoint

What it does. Turns a workflow into a REST endpoint: send a POST request with a user message, get the agent's response back. Conversation API trigger, agent, output. The default agent has no tools and a small budget (3 turns), so responses are fast and cheap.

What it needs. No connectors. You call it with an API token; see API and Integrations.

What to customize. Give the agent tools (knowledge base search makes it a RAG API) and raise the budget if you do.

Meeting Vault

What it does. Captures meeting notes into your knowledge base the moment a meeting ends. Point your meeting tool (Granola, Fireflies, Zoom, or any Zapier or webhook automation) at the workflow's API endpoint. When a summary or transcript arrives, a Meeting Librarian agent structures it into a consistent markdown note (one-line summary, attendees, key points, decisions, action items, follow-ups) and files it with knowledge.note. A Vault node on the canvas binds the "Meetings" folder of your knowledge base, so every note lands there and is indexed for retrieval. It works like a note vault your agents write for you: every chatbot and agent in the project can recall past meetings.

What it needs. No connectors. Copy the workflow's API endpoint from API and Integrations, Workflows, into your meeting tool's automation.

What to customize. The note structure in the librarian's system prompt, and the vault folder name on the Vault node.

Google Reviews Responder

What it does. Monitors new Google Business reviews and posts a professional response to each one. A Google Reviews trigger receives the review, the agent composes a reply (thankful for positive reviews, apologetic and constructive for negative ones), and an output node posts it back.

What it needs. A google_reviews connector.

What to customize. The response voice in the agent's prompt. Many businesses also add a Human In The Loop node before the reply is posted, so a person approves responses to negative reviews; see connecting nodes.

Data recipes

Two recipes form an analytics pipeline over collections, the workspace's structured datasets. They are the pattern behind scheduled ingestion and AI analysis.

Sync data source to a collection

What it does. The ingestion half. On a schedule (every 30 minutes by default), a Data Sync node pulls rows from a source table on a connector and writes them straight into a collection. No agent runs, so a cycle costs source API quota only, with no AI cost. The template syncs daily price history from Financial Market Prep into an ohlcv_daily collection and keeps a rolling 30-day window. You choose which companies to sync three ways in the node inspector: type tickers, tick an index (S&P 500, Nasdaq, Dow), or point it at a Google Sheet tab holding the list.

What it needs. A financial_market_prep connector (or another data source connector), and the ohlcv_daily collection, which the recipe creates for you.

What to customize. The source table (the inspector lists every table the source offers), the symbols, the schedule cadence, and the retention window. Use one workflow per dataset so each feed keeps its own cadence.

Analyze a collection with AI and store the results

What it does. The analysis half, and the pattern behind a strategy. Once a day, a Collection node reads the synced data, an Analyst agent applies your prompt to it (the template scores each symbol from 0 to 1 with a one-sentence rationale), and a second Collection node bound in write mode receives the results. The output collection is queryable and chartable.

What it needs. No connectors. Two collections, market_data (filled by a sync workflow) and analytics_results, both created by the recipe.

What to customize. The analysis prompt (this is your strategy), the schedule, and the collections it reads and writes. Because collections are shared across projects, this workflow can read whatever an ingestion project synced. Build one workflow per strategy, each on its own cadence.

After installing

Every installed recipe is a plain workflow. From here you can:

• Reshape it on the canvas or ask the AI Assistant for changes in plain language.
• Add channels and tools in Workflow Settings.
• Check what each attached tool needs in the Agent tool reference.
• Connect the accounts it depends on under Connectors.