### Answer There are known differences in how Flex Mode handles tool calls compared to standard (rigid) conversation flow mode. [(1) ]() Key differences in Flex Mode for tools: - There is no "Speak After Execution" setting in Flex Mode—the agent will always speak after function execution. - There is no "waitForResult" setting in Flex Mode—the agent will always wait for the function to complete (similar to Single Prompt agent behavior) . Known issues with MCP tools in Flex Mode: - Other users have reported that Flex Mode can handle tool orchestration differently than standard modes, which may cause it to bypass expected behavior during tool calls. [(2) ]() - There's also a reported issue where MCP tool response variables may not be extracted properly—the MCP server may return a stringified JSON response (e.g.,

"{\"customer_id\": 12345}"

) rather than a parsed object, which could affect how the tool result is processed. [(3) ]() Recommendations: 1. Verify your MCP server is returning properly formatted (non-stringified) JSON responses. 2. If using Flex Mode with more than 20 nodes, performance may degrade and hallucination risk increases—consider splitting into smaller components. [(1) ]() 3. Only use Prompt edges, avoid Equation edges as the LLM can have difficulty interpreting equation conditions in Flex Mode. This appears to be a known limitation/issue with Flex Mode and MCP tool integration. You may want to reach out to Retell support directly for further assistance on this specific behavior.

​ To learn more, see the sources I considered: [[1] Flex Mode](https://docs.retellai.com/build/conversation-flow/flex-mode) [[2] Flex Mode not using mocks as function call results](https://discord.com/channels/1186424376300679209/1455850410697953435) [[3] mcp tool variables not extracted](https://discord.com/channels/1186424376300679209/1426340769924382791) ​

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