Tools are the “action layer” for FlowbotAI voice agents.
They let an agent do more than talk — it can look things up, update systems,
and trigger workflows while keeping the conversation natural and on-track.
In plain terms... A tool is a function your agent is allowed to call during a
live interaction. The agent decides when to use it, fills in the required
inputs (parameters), and then uses the returned result to respond to the
caller.
You’ll also hear tools referred to as “function calling.”
Same idea — FlowbotAI exposes functions to the model so it can perform actions.
Below are practical patterns partners commonly implement
with FlowbotAI tools:
- Communicate with the outside world: pull
real-time info (weather, store hours, local events), send emails/texts, or
create calendar events — all while the caller stays on the line.
- Order and shipment status: look up an order by
phone number, order ID, or email; confirm backorders; provide shipment ETA; and
optionally send a follow-up link by email or SMS.
- Knowledge-assisted support: query a curated
knowledge base for product docs, policies, or troubleshooting steps when the
agent needs accurate, consistent answers.
- Create a support case: open a ticket with the
right fields pre-filled (caller identity, summary, urgency, device details), so
a human can pick up without re-asking basic questions.
- Escalate or transfer: hand off to a live agent,
route to the right queue, or transfer to a department with context gathered and
summarized.
- End the call gracefully: wrap up after a
successful resolution, or terminate after inactivity (with a polite warning).
FlowbotAI supports built-in tools (ready-to-use), custom
tools (your own integrations) and workflow integrations (via custom apps for Zapier, Make.com and n8n). Most deployments use at least one of these options.
Built-in tools cover common voice-agent needs without
requiring you to build anything. Examples include:
- queryCorpus: retrieve relevant information from
an existing knowledge collection (your support docs, policies, FAQs, etc.).
- warmTransfer: aka supervised transfer, transfer a call to a human operator and play a summary of the call before completing the transfer.
- coldTransfer: aka blind transfer, immediately transfer a call to a human operator.
- hangUp: programmatically end the call when the
interaction is complete or needs to stop.
Custom tools let your agent interact with your systems —
CRMs, ticketing platforms, billing, order databases, and more. Anything you can
expose as a function can become a tool.
Examples of custom tools partners build:
- Look up a customer record and return plan
details, recent invoices, or service status.
- Create or update a ticket in a helpdesk system
with structured fields (priority, category, notes).
- Take a payment or process a refund (with careful
authentication and confirmation steps).
- Trigger a workflow in a third-party automation
platform (e.g., send a Slack alert, start an approval flow).
- Send a post-call summary email to the caller or
an internal distribution list.
When the agent decides to use a tool, FlowbotAI executes the
tool and returns the result to the agent in real time. This enables live
interactions with external systems, databases, and APIs without you having to
manually “wire up” the call flow.
Because the caller is waiting, tools should be designed to
respond quickly. Long-running operations can make the conversation feel broken,
so keep responses fast or use async patterns where available in your
deployment.
Most tools fall into one of two implementation styles. Your
choice depends on where the logic needs to run and what you’re integrating
with.
Server-based (HTTP) tools
- Most common approach: FlowbotAI calls an HTTP
endpoint you control.
- Best for integrating with back-office systems,
databases, and internal services.
- Easier to secure and audit because requests are
centralized.
Client-side tools (advanced) - COMING SOON!
- Runs directly inside your client application
(for example, a web or mobile app embedding a FlowbotAI experience).
- Useful when you need access to device-local
capabilities or client-only context.
- Typically requires an SDK-based integration and
is less common for standard phone-based deployments.
FlowbotAI tools can be set up for reuse across agents and
calls, or defined for a single session during testing and iteration.
Reusable (durable) tools - (default)
- Created once and reused across agents and calls.
- Best fit for production deployments and shared
integrations.
- Easier to manage over time because updates apply
everywhere the tool is used.
Session-defined (temporary) tools - COMING SOON!
- Defined for a single test or runtime session
(useful for rapid experimentation).
- Great for prototyping, but not usually the best
choice for long-term partner deployments.
- If your environment doesn’t expose temporary
tool configuration in the portal, treat this as an advanced integration
pattern.
A simple path for first-time setup:
- Choose the outcome you want (e.g., “look up an order,” “create a ticket,” “transfer to billing”).
- Select any built-in tools your agent needs for call control or knowledge lookup.
- Add a custom tool for each external system you want to connect (CRM, helpdesk, order system).
- Update your agent prompt with clear rules for when to call each tool and what to do with the results.
- Test with real-world phrases and edge cases (missing data, partial matches, caller refuses verification).
Before you launch a tool-enabled agent to customers, make sure:
- The tool responds quickly and reliably (aim for a few seconds or less for most calls).
- Inputs are explicit and validated (required fields, formats, and safe defaults).
- The tool returns only what the agent needs (avoid dumping large records into the conversation).
- Sensitive actions require verification and confirmation (payments, account changes, refunds).
- Your prompt includes clear rules for when to call the tool and how to summarize results for the caller.
- You have a graceful fallback message for timeouts, errors, or “no match found” outcomes.