Retell AI MCP Server
Provides access to Deepgram voices for use in voice agents.
Provides access to ElevenLabs voices for use in voice agents.
Provides access to OpenAI voices for use in voice agents.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Retell AI MCP Server list my voice agents and their details"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Retell AI - Model Context Protocol (MCP) Server
A fully-featured, official-grade Model Context Protocol (MCP) server that provides complete programmatic access to the Retell AI platform.
This MCP server achieves 100% parity with the Retell Node.js SDK, exposing 60 specific tools allowing LLM-powered assistants (like Claude Desktop, Cursor, and Codex) to autonomously build, configure, and operate end-to-end voice and text agents.
Capabilities
With this server, your AI assistant can autonomously:
Design Agents: Create and configure LLM response engines, Conversation Flows, and Chat Agents.
Provide Memory: Create Knowledge Bases, upload content, and link them to agents for RAG.
Manage Telephony: Purchase, import, and configure U.S. or international phone numbers.
Trigger outbound calls & test: Initiate Web calls, Phone calls, SMS chats, and batch tests.
Monitor: Retrieve transcripts, real-time metrics, live call concurrency, and latency scores.
Everything available in the official SDK is available as an MCP Tool.
Related MCP server: Retell Universal MCP Server
Installation
Prerequisites
Node.js 18+
A Retell AI API Key (from the Retell Dashboard)
Setup
Clone the repository and install dependencies:
git clone https://github.com/sunnysingh100/retell-mcp-server.git
cd retell-mcp-server
npm install
npm run buildConfiguration
To use this with an MCP client (such as Claude Desktop or Codex), you must configure it via your client's settings.
Claude Desktop
Edit your claude_desktop_config.json:
{
"mcpServers": {
"retell-ai": {
"command": "node",
"args": [
"/absolute/path/to/retell-mcp-server/dist/index.js"
],
"env": {
"RETELL_API_KEY": "your_retell_api_key_here"
}
}
}
}Generic Client
Just run the built script with the environment variable set. The server operates over stdio via the standard MCP JSON-RPC protocol.
RETELL_API_KEY="your_api_key" node dist/index.jsAvailable Tools (60 Total)
The server supports all 14 domain resources of the Retell SDK.
🎙 Voices
retell_list_voices: List available voices across providers (ElevenLabs, OpenAI, Deepgram, etc).
🧠 LLMs & Knowledge Bases
retell_list_llms/retell_get_llm/retell_create_llm/retell_update_llm/retell_delete_llmretell_list_knowledge_bases/retell_get_knowledge_base/retell_create_knowledge_base/retell_delete_knowledge_baseretell_add_knowledge_base_sources/retell_delete_knowledge_base_source
🤖 Voice Agents
retell_list_agents/retell_get_agent/retell_create_agent/retell_update_agent/retell_delete_agentretell_get_agent_versions/retell_publish_agent
📱 Phone Numbers
retell_list_phone_numbers/retell_get_phone_number/retell_create_phone_numberretell_update_phone_number/retell_delete_phone_number/retell_import_phone_number
📞 Voice Calls
retell_list_calls/retell_get_call/retell_update_call/retell_delete_callInitiate:
retell_create_phone_call/retell_create_web_call/retell_register_phone_call
💬 Chat Agents & Sessions
Agents:
retell_list_chat_agents/retell_get_chat_agent/retell_create_chat_agent/retell_update_chat_agent/retell_delete_chat_agent/retell_get_chat_agent_versions/retell_publish_chat_agentSessions:
retell_list_chats/retell_get_chat/retell_create_chat/retell_update_chat/retell_create_chat_completion/retell_create_sms_chat/retell_end_chat
🔀 Conversation Flows (Node-based scripting)
Flows:
retell_list_conversation_flows/retell_get_conversation_flow/retell_create_conversation_flow/retell_update_conversation_flow/retell_delete_conversation_flowComponents:
list,get,create,update,deletefor reusable conversation flow components.
⚙️ Utilities
retell_create_batch_call/retell_create_batch_testretell_get_concurrencyretell_get_mcp_tools
Architectural Notes
Input Validation: Every tool validates its inputs dynamically using strict
zodschemas. Invalid arguments return clean, detailed error messages.Type Safety: The entire codebase is thoroughly typed against the
retell-sdk, eliminatinganycasting and masking TypeScript errors.Standard Protocol Compliance: All expected errors generate an MCP protocol
isError: truepayload, ensuring the LLM client correctly handles API rejections without crashing.
Available Tools
60 toolsretell_add_knowledge_base_sourcesA
Add new text or URL sources to an existing knowledge base.
Inputs: knowledge_base_id (required), knowledge_base_urls, knowledge_base_texts.
retell_create_agentA
Create a new Retell AI voice agent. Requires a voice_id and one response engine: llm_id (Retell LLM), llm_websocket_url (Custom LLM), or conversation_flow_id (Conversation Flow).
Inputs: voice_id (required), llm_id, language, agent_name, +7 more.
retell_create_batch_callA
Schedule a batch of outbound phone calls. Each task specifies a destination number and optional per-call agent overrides and dynamic variables.
Inputs: tasks (required), from_number (required), name, trigger_timestamp.
retell_create_batch_testA
Run a batch of automated test cases against a Retell LLM or Conversation Flow response engine.
Inputs: response_engine (required), test_case_definition_ids (required), reserved_concurrency.
retell_create_chatB
Start a new chat session with a chat agent.
Inputs: agent_id (required), metadata, agent_version, retell_llm_dynamic_variables.
retell_create_chat_agentA
Create a new chat (text/messaging) agent. Requires a response_engine (llm_id, llm_websocket_url, or conversation_flow_id).
Inputs: llm_id, language, agent_name, webhook_url, +2 more.
retell_create_chat_completionB
Send a user message to an ongoing chat session and get the agent's response.
Inputs: chat_id (required), content (required).
retell_create_conversation_flowA
Create a new Conversation Flow. A Conversation Flow is a node-graph that controls agent dialogue with conversation nodes, function nodes, branch nodes, and end nodes connected by edges.
Inputs: nodes (required), model_choice (required), start_speaker (required), tools, +4 more.
retell_create_conversation_flow_componentA
Create a new shared Conversation Flow Component. Components are reusable sub-flows that can be embedded across multiple Conversation Flows.
Inputs: name (required), nodes (required), tools, start_node_id.
retell_create_knowledge_baseA
Create a new knowledge base for RAG. Provide plain-text sources, URLs to crawl, or both.
Inputs: knowledge_base_name (required), knowledge_base_urls, knowledge_base_texts.
retell_create_llmB
Create a new Retell LLM configuration to attach to agents. Define system prompt, model, begin message, and tools.
Inputs: model, begin_message, general_tools, general_prompt, +1 more.
retell_create_phone_callB
Initiate an outbound phone call via a Retell AI agent. Requires a purchased from_number and a destination to_number.
Inputs: to_number (required), from_number (required), metadata, override_agent_id, +1 more.
retell_create_phone_numberA
Purchase a new US phone number and optionally bind it to an agent.
Inputs: nickname, area_code, inbound_agent_id, outbound_agent_id.
retell_create_sms_chatB
Start an outbound SMS chat conversation using a phone number with SMS capability.
Inputs: to_number (required), from_number (required), metadata, override_agent_id, +1 more.
retell_create_web_callA
Create a browser-based web call with a Retell AI agent. Returns an access_token for the Retell Web SDK.
Inputs: agent_id (required), metadata, retell_llm_dynamic_variables.
retell_delete_agentA
Permanently delete a Retell AI agent by its ID.
Inputs: agent_id (required).
retell_delete_callB
Delete a call record from your Retell AI account.
Inputs: call_id (required).
retell_delete_chat_agentA
Permanently delete a chat agent by its ID.
Inputs: agent_id (required).
retell_delete_conversation_flowA
Permanently delete a Conversation Flow and all its versions.
Inputs: conversation_flow_id (required).
retell_delete_conversation_flow_componentA
Delete a shared Conversation Flow Component. When deleted, local copies are automatically created in all Conversation Flows that use it.
Inputs: conversation_flow_component_id (required).
retell_delete_knowledge_baseB
Delete a knowledge base from your Retell AI account.
Inputs: knowledge_base_id (required).
retell_delete_knowledge_base_sourceA
Remove a specific source from an existing knowledge base by its source ID.
Inputs: source_id (required), knowledge_base_id (required).
retell_delete_llmB
Delete a Retell LLM configuration by its ID.
Inputs: llm_id (required).
retell_delete_phone_numberB
Release (delete) a phone number from your Retell AI account.
Inputs: phone_number (required).
retell_end_chatB
End an ongoing chat session.
Inputs: chat_id (required).
retell_get_agentA
Get detailed information about a specific Retell AI agent by its ID.
Inputs: agent_id (required).
retell_get_agent_versionsA
Get all published versions of a Retell AI voice agent.
Inputs: agent_id (required).
retell_get_callA
Get full details of a specific call including transcript, recording URL, latency, and cost.
Inputs: call_id (required).
retell_get_chatA
Get details and transcript of a specific chat session.
Inputs: chat_id (required).
retell_get_chat_agentB
Get details of a specific chat agent by its ID.
Inputs: agent_id (required).
retell_get_chat_agent_versionsB
Get all published versions of a chat agent.
Inputs: agent_id (required).
retell_get_concurrencyA
Get the current live call concurrency and concurrency limits for your Retell organization.
Inputs: no named parameters.
retell_get_conversation_flowA
Get full details of a specific Conversation Flow including its nodes, edges, and tools.
Inputs: conversation_flow_id (required).
retell_get_conversation_flow_componentA
Get full details of a specific shared Conversation Flow Component.
Inputs: conversation_flow_component_id (required).
retell_get_knowledge_baseA
Get details about a specific knowledge base.
Inputs: knowledge_base_id (required).
retell_get_llmA
Get details of a specific Retell LLM configuration.
Inputs: llm_id (required).
retell_get_mcp_toolsA
Get the list of MCP tools available from a specific MCP server configured on a Retell agent.
Inputs: mcp_id (required), agent_id (required), version.
retell_get_phone_numberA
Get details about a specific phone number including its associated agent.
Inputs: phone_number (required).
retell_import_phone_numberA
Import a phone number from your own custom telephony (SIP trunk) and optionally bind agents to it.
Inputs: phone_number (required), termination_uri (required), nickname, inbound_agent_id, +3 more.
retell_list_agentsA
List all Retell AI voice agents in your account. Returns agent IDs, names, voice settings, and configuration.
Inputs: no named parameters.
retell_list_callsA
List calls in your Retell AI account. Supports filtering by agent ID, status, and date range. Supports pagination via pagination_key.
Inputs: limit, sort_order, pagination_key, filter_criteria.
retell_list_chat_agentsA
List all chat (text/messaging) agents in your Retell AI account.
Inputs: no named parameters.
retell_list_chatsA
List all chat sessions in your Retell AI account.
Inputs: no named parameters.
retell_list_conversation_flow_componentsA
List all shared Conversation Flow Components in your account. Components are reusable sub-flows that can be embedded in multiple Conversation Flows, saving development time and ensuring consistency.
Inputs: no named parameters.
retell_list_conversation_flowsA
List all Conversation Flows in your Retell AI account. Conversation Flows are node-based agent scripts with fine-grained control over dialogue, branching, and function execution.
Inputs: no named parameters.
retell_list_knowledge_basesB
List all knowledge bases in your Retell AI account.
Inputs: no named parameters.
retell_list_llmsA
List all Retell LLM configurations in your account.
Inputs: no named parameters.
retell_list_phone_numbersA
List all phone numbers in your Retell AI account.
Inputs: no named parameters.
retell_list_voicesA
List all available voices for Retell AI agents, including provider, gender, accent, and preview audio URLs.
Inputs: no named parameters.
retell_publish_agentA
Publish the latest draft version of a voice agent, creating a new immutable version and starting a fresh draft.
Inputs: agent_id (required).
retell_publish_chat_agentA
Publish the latest draft of a chat agent, creating a new immutable version.
Inputs: agent_id (required).
retell_register_phone_callA
Register a phone call for custom telephony (bring-your-own telephony). Returns call_id and access_token for the audio websocket.
Inputs: agent_id (required), metadata, direction, to_number, +2 more.
retell_update_agentB
Update an existing Retell AI agent's configuration, including switching to a different response engine type (Retell LLM, Custom LLM, or Conversation Flow).
Inputs: agent_id (required), llm_id, language, voice_id, +8 more.
retell_update_callB
Update metadata or data storage settings for an existing call.
Inputs: call_id (required), metadata, data_storage_setting.
retell_update_chatC
Update metadata or data storage settings for an ongoing chat session.
Inputs: chat_id (required), metadata, data_storage_setting, override_dynamic_variables.
retell_update_chat_agentB
Update an existing chat agent's configuration.
Inputs: agent_id (required), llm_id, language, agent_name, +3 more.
retell_update_conversation_flowB
Update an existing Conversation Flow's nodes, edges, model, prompt, or tools.
Inputs: conversation_flow_id (required), nodes, tools, model_choice, +5 more.
retell_update_conversation_flow_componentB
Update an existing shared Conversation Flow Component's name, nodes, or tools.
Inputs: conversation_flow_component_id (required), name, nodes, tools, +1 more.
retell_update_llmA
Update an existing Retell LLM configuration's prompt, model, or tools.
Inputs: llm_id (required), model, begin_message, general_prompt, +1 more.
retell_update_phone_numberB
Update the agent binding or settings for an existing phone number.
Inputs: phone_number (required), nickname, inbound_agent_id, outbound_agent_id.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 60 tools
Each tool targets a distinct resource type (phone numbers, agents, calls, LLMs, knowledge bases, conversation flows, chat agents, chats, batch, concurrency, MCP tools) with clear operations. Potential overlaps like retell_create_phone_call vs retell_create_web_call are differentiated by channel, and retell_list_agents vs retell_list_chat_agents are separated by chat/voice context.
All tools follow the retell_verb_noun pattern with consistent verbs (list, get, create, update, delete). Exceptions like retell_add_knowledge_base_sources and retell_publish_agent are conventional and still follow the pattern. No mixing of conventions or ambiguous verbs.
60 tools is well above the typical 3-15 range, but the server covers a broad platform API with many resource types. Each tool maps to a distinct operation, making the count justifiable. However, the sheer number may overwhelm agents, so it is not ideal.
The tool set provides comprehensive CRUD for all major resources, including lifecycle management for agents (versions, publish) and calls (list, get, create, update, delete, register). Minor gaps exist, such as no direct update for knowledge base metadata and no delete chat session, but these are not critical dead ends.
Maintenance
Related MCP Connectors
Create and manage DialNexa AI voice agents, campaigns, calls, phone numbers, workflows, and more.
Build, test, deploy and run AI phone agents: agents, numbers, calls, tests, knowledge, campaigns.
Manage CallKaro voice agents, phone numbers, call queues, and place single or batch AI calls.
Give your AI assistant a real phone line. Place and end real phone calls with AI voice agents, read call transcripts, run a live two-way interpreter between two people who share no language (31 languages, browser link or phone), create and edit voice agents, and search, buy and bind phone numbers in 21 countries. OAuth 2.1 with PKCE — the model never sees your API key; a read-only scope is available. Pay as you go from $0.10/min, $5 free credit for new accounts.
Related MCP Servers
- AlicenseBqualityDmaintenanceEnables interaction with Retell AI's voice and chat agent platform. Build, deploy, and manage AI phone agents, configure conversation flows, handle calls/chats, and manage phone numbers through natural language.52MIT
- AlicenseNot gradedqualityDmaintenanceProvides a standardized interface for interacting with Retell's tools and services through a unified API.1MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to manage RetellAI voice services including calls, agents, phone numbers, and voices through natural language.ISC
- AlicenseNot gradedqualityBmaintenanceEnables building, publishing, and operating CallDeskTech voice agents, including conversation flows, knowledge bases, phone routing, calls, batch calls, and webhooks.37 npmMIT