Skip to content
EN
Open app

For AI assistants: this page as plain markdown is at https://docs.voicelogica.ai/raw/getting-started/your-first-agent.md, and https://docs.voicelogica.ai/llms.txt lists the docs by topic, one file each, small enough to read whole. Read those rather than this HTML page.

Your First Agent

12 min read
Copy as Markdown
View as MarkdownOpen the raw .md file.md

Sign-up already created your first agent from your company details and handed you a test line. This page is what you do next: open that agent, shape how it talks, try it, and review the call.

The agent name comes from onboarding (for example Agent Of Voice Logica Docs when the company is Voice Logica Docs). You can rename it later in Instructions - you do not need to recreate it.

In the sidebar, open AI agents. The page heading is My Agents.

My Agents list with the onboarding agent
first-agent-1Expand

What you see:

  • Create agent - adds another agent if your plan allows it (often shows a lock when you are already at the limit)
  • Copy agent - duplicates an existing agent when your plan allows it
  • Add filter and Search agents… - narrow the list
  • One agent card from onboarding - name, how recently it was touched, a short ID chip, View calls, and quick actions on the card

Click the agent card to open the editor.

Click Create agent. If your plan includes only one agent, Agent limit reached appears with Your plan includes 1 agent - upgrade to add more. Choose View subscription to upgrade, or Cancel and keep customizing the agent you have.

Agent limit reached dialog
first-agent-5Expand

Resellers, and Voice Logica account managers, can send an agent to another company from Copy agent:

  1. Pick the agent, tick Transfer to another company.
  2. Choose Copy (the agent stays here too) or Move (it leaves this company).
  3. Pick the company: the list opens with every company you can send it to, A to Z. Type to narrow it.

Move is refused while something still uses the agent here (a number or extension, an open campaign, a chat channel, the website widget, a PBX connector, a lead form). The message names each one: point them to another agent, then move it.

The editor header shows the agent name, a green Live / v1 badge, and Prompt Based - editing the live agent.

Four header controls sit above the tabs:

Required
Readiness checks. Opens Not ready for calls with required items and warnings before callers can reach the agent.
Edit with AInarrow: AI
Opens Claude / ChatGPT against this agent (needs an AI connection).
Live on
Which channels this agent answers on. 0 of 2 channels means Try this agent can work while customers still cannot reach it.
Try this agentnarrow: Try
Two ways to hear the agent before real callers do.
Required checklist Not ready for calls
first-agent-requiredExpand

Required can list blockers such as Telephony (callers cannot reach the agent yet), plus warnings like tools mentioned in the prompt that are not enabled, or AI disclosure missing from the greeting. In the editor there is no separate publish step - saving updates the live agent. To try a change before callers hear it, make it from a chat with Claude or ChatGPT: it is saved as a new numbered version (v2, v3 and so on) while the live agent keeps answering. You test it (call the dial-in PIN with * and the version number added, for example *2 for v2, or run test scenarios), then publish it. See Build a reliable phone agent.

For the first week you only need Instructions and Voice & Language. The other rail areas are the Agent Configuration reference - open that after a good test call.

Instructions tab with Basic Info and Greeting
first-agent-2Expand
IdentityTHIS PAGE
InstructionsVoice & Language

Three sections:

  1. Basic Info

    Agent nameREQUIRED
    The name callers and your team see in chats and call logs. Onboarding fills this from your company details.
    Internal dial extensionASSIGNED410
    Dial this from any internal phone. The system assigns it - it cannot be changed.
    Advanced - Reply stylecollapsed by default
    Open it only if you need to tune how replies are phrased.
  2. Greeting

    AI disclosure
    When on, callers are told they are speaking to an AI if your greeting does not already say so. Voice Logica surfaces Article 50(1) and 50(5) of the EU AI Act here. If the toggle is off and the greeting does not disclose AI, you (the deployer) may not be compliant.
    Welcome message
    The first thing callers hear. Use Insert greeting for a starting line. Keep it short.
    Advanced - Greetingcollapsed by default
    Dynamic greeting: the AI writes a different greeting each call instead of the fixed text. Allow interruption: callers can talk over the welcome message and the agent stops to listen. Off by default: the welcome message always plays to the end, and anything the caller says during it is ignored. Barge-in for the rest of the call is set separately under Call handling - Advanced. Skip key for the welcome: pick a key (for example 9) and a caller who presses it stops the welcome message, recording notice included, and goes straight to the agent. Off by default.
  3. AI personality & behavior

    This is the system prompt - the biggest lever on call quality. Cover:

    • The agent’s role and the company it represents
    • Tone (professional, warm, formal, casual)
    • What to collect, what to refuse, when to transfer or hang up
    • How to handle common questions

    Controls above the prompt:

    Prompt history
    Restore an earlier version.
    Insert variable
    Live values such as the caller’s number or today’s date. Use {{variable_name}}.
    Word pronunciation
    Teach brand names and acronyms.
    Improve with AI
    Refine the prompt in place.

    A character counter sits under the prompt (plan-dependent, for example 5,121 / 10,000). If you run out of room, move reference material into a knowledge base instead of stuffing the prompt.

Pick a primary language and a voice first. Add secondary languages later for callers who prefer them.

  • Primary Language & Voice - Primary language is the language the agent starts every call in. Voice is the speaking voice; Preview plays a sample, Change voice opens the catalog. Advanced - Voice tuning is collapsed by default.
  • Voice model (ElevenLabs voices, in Advanced - Voice tuning) - Expressive sounds the most human and acts out emotion from the text and audio tags; you set only stability (around 0.5 suits most calls) and similarity. Most options gives the fastest replies and every voice setting (speed, stability, similarity, style, speaker boost). Platform default follows the platform setting. Make a test call after you switch, to hear the difference. All voice settings: Voices.
  • Secondary Languages - Add language so the agent can switch when a caller prefers another language. It switches at once when the caller asks for it (“can we speak English?”). When the caller simply speaks another language, it first asks whether they want to continue in it, so a noisy line or a single foreign word never flips the call.
  • Call Background Audio - optional ambient sound under the whole call: office, call center or coffee shop. Keep the volume low (about 10-20%): loud ambience covers the voice and makes the agent mishear callers. It plays in every language. There is no music under the conversation; music only plays while a call is being transferred.
  • Accuracy - Custom vocabulary for uncommon names and terms. Type a word and press Enter or comma.
Voice and Language - primary language and voice
first-agent-voiceExpand

Click Try this agent in the editor header. The panel title is Try this agent. Two ways to hear it before callers do.

Try this agent panel with browser and phone options
first-agent-3Expand
  1. Talk to it in your browser

    Uses your mic. Nothing goes live. This is the fastest loop while you edit the greeting or prompt.

  2. Call it from your phone

    Dial the test number, then enter the PIN (with #). Dial and Copy are on the panel.

    This is the same test line you received at the end of sign-up.

  3. Review the conversation

    Open Calls in the sidebar (or View calls on the agent card). Use the History tab (it may show a count) and Scheduled. Each row shows the caller, channel, agent, date, duration, and status (for example Answered).

    Calls History with an answered test call
    first-agent-4Expand

    Open a call to listen to the recording and read the transcript, then go back to Instructions or Voice & Language and adjust.

A text-only agent (for example a website chat) does not need the test line. Uncheck Calls with the PIN on the test line at the bottom of the panel and Save: dialing the PIN then no longer reaches this agent. The PIN is kept, so checking the box again brings it back unchanged. To also hide the voice button in the website chat, see Text chat only.

Try this agent is a private rehearsal line. Real customers reach the agent through a channel. Open Live on in the header to see Channels.

Live on Channels panel
first-agent-live-onExpand
  1. Check 0 of 2 channels and the status line (for example Not reachable on any channel yet).
  2. Turn on Website Widget, or use Telephony → Connect to attach a number - see Voice Logica Phone Numbers.
  3. Keep a paid subscription so you have minutes for real traffic.

Then put the number on your website, email signatures, and listings.

1

Open the onboarding agent - do not create a second one until you need it.

2

Rewrite the greeting - callers should know who they reached in the first few seconds, including the AI disclosure if the law in your region requires it.

3

Write a specific personality prompt - role, tone, what to collect, when to transfer.

4

Pick language and voice - preview the sample before you save.

5

Try it, then open Calls - browser or phone, then listen and iterate.

Why can't I find a Create agent wizard for my first agent?

The wizard is no longer part of the new-user path: sign-up already created your first agent when you finished the business form. Open AI agents / My Agents and click the agent card that is already there.

Why does Create agent say Agent limit reached?

Your plan includes a fixed number of agents (often 1 on Demo and the first paid tier), so this is a plan limit, not a product bug. Click View subscription to upgrade, or keep shaping the existing agent.

What does the Required button on my agent mean?

Required opens the Not ready for calls list. Required items block real callers (for example Telephony not connected), while warnings point out mismatches such as tools mentioned in the prompt that are still off, or AI disclosure missing from the greeting. Fix each item and the Required chip clears.

Do I have to turn on AI disclosure for my agent?

Under Article 50(1) and 50(5) of the EU AI Act, callers must be told at the first interaction that they are speaking with AI. If your greeting already says so, you can leave the AI disclosure toggle off; if it does not, turn it on. You are the deployer: Voice Logica surfaces the reminder, but compliance is yours.

Where is the test line to try my agent?

The test line is under Try this agent in the agent editor header, not a permanent banner. Open that panel and choose Talk to it in your browser or Call it from your phone (number + PIN).

Do I need to publish my agent after I save changes?

No. Saving in the editor updates the live agent (Live / v1), and the footer reminds you that saved changes go to callers. Only a change you save as a version from a chat with Claude or ChatGPT (destination: "version") waits: publish it with versions_execute (action "publish_agent_version", with its versionId or versionNumber) once you have tested it.

DOCUMENTATION
LIVE CHAT
Available in your dashboard via the Support button