Skip to content

Add a tool manually

Last updated: 2026-06-30

This article shows how to create a custom tool by filling in its configuration in the Dromlik admin console.

Before you begin

From the third-party API provider, gather:

ItemDescription
Endpoint URLThe API request address — must use HTTPS.
Request methodGET, POST, PUT, PATCH, or DELETE.
ParametersPath, header, query, and/or body parameters the request needs.
CredentialsStatic credentials only — API key, Basic Auth username/password, or a static Bearer token. OAuth 2.0 is not supported.

Procedure

  1. 1

    Open the Tools page

    Sign in to the Dromlik admin console and go to AI > AI Receptionist > Tools.

  2. 2

    Click Add

    Click Add. You can switch between Form Editor and JSON Editor from the top-right of the panel at any time.

  3. 3

    Configure basic information

    SettingDescription
    NameUnique tool name. Only letters, numbers, _, and -.
    Request MethodThe HTTP method the API expects.
    URLAPI endpoint, starting with https://. Up to 5 path parameters in {braces} are supported.
    RemarkOptional internal note about the tool's purpose.
    DescriptionNatural-language instruction telling the receptionist when and how to call the tool. Be specific — this is what the model uses to decide.
  4. 4

    Configure execution rules

    SettingDescription
    Disable Interruptions During Tool ExecutionCallers can't interrupt while the tool runs.
    Play Prompt Before Executing ToolThe receptionist always speaks a filler ("Let me check that…") before the call. When off, it decides based on recent latency.
    Execution Mode
    • Execute Immediately on LLM Request — runs as soon as the model requests it.
    • Execute After Voice Prompt — runs after the receptionist speaks a prompt.
    • Async — runs in the background; the receptionist doesn't wait. Use for fire-and-forget actions like sending an email.
    Tool Execution Sound EffectAmbient sound played during the call: None, Typing, or Music 1–4. Not available in Async mode.
    Sound Effect Playback RulesPlay only After Voice Prompt or Always play.
    Response Timeout (s)Between 5 and 120 seconds.
  5. 5

    Configure request parameters

    Define the four kinds of parameters the API needs:

    Path parameters

    Auto-detected from the URL placeholders (e.g. /orders/{order_id}). For each, choose a Data Type (String / Boolean / Integer / Number) and a Value Resolution Logic:

    • Constant Value — fixed value you specify.
    • Generated by LLM — the receptionist extracts the value from the conversation using your Description.

    For LLM-generated string parameters, you may add up to 10 Optional Values the receptionist must pick from.

    Headers

    Up to 20 static headers. Use for static auth — e.g. Authorization: Bearer <key>. Dynamic values are not supported.

    Query parameters

    Up to 10. Same value-resolution choices as path parameters, plus a Required flag — when set, the receptionist will keep asking the caller until it has a value.

    Body parameters

    Available only for POST, PUT, and PATCH. Up to 30 parameters. Supports the same scalar types as above, plus:

    • Object — nested key/value group with up to 10 child parameters.
    • Array — a list of values of one declared element type.

    Include a top-level Body description telling the receptionist how to extract the whole payload from the call.

  6. 6

    Save

    Click Save. The tool appears in the tool list and is ready to be assigned to an AI Receptionist.

What to do next

Assign the tool to an AI Receptionist from its Skills tab, then place a test call to confirm the receptionist invokes it correctly.