### Answer Dynamic variables in Retell AI work differently depending on their source: Default System Variables - These are automatically provided by Retell and don't require configuration [(1) ]() . They include variables like

{{current_time}}

,

{{current_time_[timezone]}}

,

{{call_type}}

,

{{direction}}

,

{{user_number}}

, and

{{agent_number}}

. These system variables are reliably populated by Retell's infrastructure. Extracted Dynamic Variables - When you use the Extract Dynamic Variables tool, the LLM extracts values from user responses and saves them [(2) ]() . You can extract different types including Text, Number, Enum, or Boolean . Since this extraction relies on the LLM interpreting user input, there is potential for the LLM to misinterpret or incorrectly extract values. Dynamic Variables from Tool Calls - When you set response variables in MCP tools or custom functions, you extract values from the tool response and save them as dynamic variables [(3) ]() . The accuracy here depends on what your external tool returns. Handling Missing Variables - When a dynamic variable has no assigned value, it remains in its raw form with the curly braces intact [(1) ]() . For example, if

user_name

is not provided, the prompt would show

"Hello {{user_name}}, how can I help you today?"

. To minimize issues with extracted variables, it's recommended to include explicit instructions in your prompt about when to invoke extraction functions [(2) ]() . For example: "When user states his name and phone number, please extract the information by calling the `extract_user_detai

ls` function" .

​ To learn more, see the sources I considered: [[1] Dynamic Variables](https://docs.retellai.com/build/dynamic-variables) [[2] Extract Dynamic Variables](https://docs.retellai.com/build/single-multi-prompt/extract-dv) [[3] MCP](https://docs.retellai.com/build/single-multi-prompt/mcp) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Yes, you can use booking IDs from a tool call response in subsequent tool calls by extracting them as response variables [(1) ]() [(2) ]() . How to set this up: When configuring your custom function or MCP tool, you can extract values from the API response and save them as dynamic variables . For example, if your tool returns:

json
{
  "booking_ids": ["xyz", "abc", "qwe"]
}

You would set up response variables using JSON path notation to extract specific values [(3) ]() . For arrays, you can use syntax like

booking_ids[0]

to extract the first booking ID . Reliability considerations: The extracted values will be stored as dynamic variables and can be referenced later using

{{variable_name}}

syntax [(4) ]() . However, there are some important points: 1. All dynamic variable values must be strings . If your tool returns an array or complex object, you'll need to extract individual elements or handle the structure appropriately. 2. Variable extraction depends on correct JSON path mapping [(3) ]() . You need to specify the exact path to each value you want to extract. 3. Missing variables remain in raw form - if a variable isn't set, it will appear as

{{variable_name}}

with curly braces intact [(4) ]() . Best practice: Extract each booking ID as a separate dynamic variable (e.g.,

{{booking_id_1}}

,

{{booking_id_2}}

,

{{booking_id_3}}

) using the response variables feature, then reference them in subsequent tool calls [(1) ]() [(2) ]() .

​ To learn more, see the sources I considered: [[1] Integrate any system with custom function](https://docs.retellai.com/build/single-multi-prompt/custom-function) [[2] MCP](https://docs.retellai.com/build/single-multi-prompt/mcp) [[3] Code Tool](https://docs.retellai.com/build/single-multi-prompt/code-tool) [[4] Dynamic Variables](https://docs.retellai.com/build/dynamic-variables) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Based on the available documentation, dynamic variables in Retell AI are persistent throughout a call session once they are set [(1) ]() . The LLM should not "lose" dynamic variables during a call. How dynamic variables work: Dynamic variables are injected into your agent's prompts and tool descriptions using the

{{variable_name}}

syntax . Once set (either through the API, webhooks, or extracted during the call) , they remain available for the duration of that call session. For variables from tool responses: When you use the "Store Fields as Variables" feature in custom functions or MCP tools, values extracted from API responses are saved as dynamic variables [(2) ]() . These can then be referenced in subsequent tool calls using

{{variable_name}}

. Important considerations: 1. All dynamic variable values must be strings [(1) ]() [(3) ]() 2. Variables are call-scoped - they persist for the duration of the call 3. Missing variables remain in raw form - if a variable isn't set, it appears as

{{variable_name}}

with curly braces [(1) ]() Potential issue with agent transfers: There is one documented case where dynamic variables may not persist: during agent transfers. A community discussion indicates that dynamic variables set in one agent may not automatically transfer to the swapped agent [(4) ]() . However, this appears to have been resolved in that specific case. Bottom line: For normal tool-to-tool calls within the same agent during a single call session, dynamic variables

should remain reliable and not be lost by the LLM.

​ To learn more, see the sources I considered: [[1] Dynamic Variables](https://docs.retellai.com/build/dynamic-variables) [[2] Code Tool](https://docs.retellai.com/build/single-multi-prompt/code-tool) [[3] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[4] Dynamic Variables not persisted](https://discord.com/channels/1186424376300679209/1432409107393937589) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer I don't have specific information about "finetune examples" in the available sources. The sources provided cover API references for creating, updating, getting, and listing Retell LLMs [(1) ]() [(2) ]() [(3) ]() [(4) ]() , as well as deployment guidance [(5) ]() , but they don't contain details about finetune examples or whether they work for both single prompt and conversational flow agents. For information about fine-tuning capabilities and how they apply to different agent types, I'd recommend checking out these potentially useful sources: - [Create Retell LLM]() - [Update Retell LLM]() - [How to Deploy Conversational AI | Retell AI]() Could you clarify what specific aspect of fine-tuning you're interested in? That would help me provide more targeted guidance if the information becomes available.

​ To learn more, see the sources I considered: [[1] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[2] Get Retell LLM](https://docs.retellai.com/api-references/get-retell-llm) [[3] Create Retell LLM](https://docs.retellai.com/api-references/create-retell-llm) [[4] List Retell LLMs](https://docs.retellai.com/api-references/list-retell-llms) [[5] How to Deploy Conversational AI | Retell AI](https://www.retellai.com/blog/how-to-deploy-conversational-ai) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.