Retell AI MCP server for your AI

Place an OUTBOUND phone call with a Retell agent.

Create a WEB (browser/in-app) call with an agent — no phone number needed.

Permanently delete a call record (path call_id), including its transcript and recording.

Get a single call by call_id — the way to POLL an in-progress or finished call.

List / filter calls (current endpoint).

End / hang up an active call immediately.

Create a VOICE agent.

Start a text CHAT session with a chat agent.

Send an user message into an existing chat and get the agent's reply (the chat equivalent of one conversational turn).

Delete a voice agent and all its versions.

End an ongoing chat (path chat_id) so post-chat analysis runs.

Get a voice agent's full config by agent_id.

Get one chat by chat_id — its message_with_tool_calls transcript, chat_status, chat_analysis (summary/sentiment/custom data) and cost.

List voice agents (current endpoint).

List / filter chats (current endpoint).

Publish a specific draft version of an agent so it goes live for calls.

Update a voice agent.

Create a CHAT agent (text/SMS conversations, no voice).

Buy / provision a NEW phone number from Retell (Twilio or Telnyx) and assign agents to it.

Release / delete a phone number.

Get a chat agent's config by agent_id.

Get a knowledge base by id (path knowledge_base_id) — its name, status, and the list of sources (each with its source_id) so you can delete one with delete_knowledge_base_source.

Get one phone number's config by the number itself (E.164, path param, e.g.

Import an EXISTING number you already own (your SIP trunk / carrier) into Retell.

List all chat agents in the account.

List all knowledge bases in the account with their ids and statuses.

List all phone numbers in the account with their agent assignments (current endpoint).

Update a phone number's config (path = the number in E.164) — reassign agents, change webhook url, SIP auth, or fallback number.

Create a Conversation Flow — a node/branch-based response engine (visual-flow brain).

Create a Retell LLM — a prompt-based response engine (the agent's brain).

Delete a conversation flow.

Delete a Retell LLM.

Get a conversation flow by id (path conversation_flow_id), including its full node graph — useful as a template before authoring or updating one.

Get a Retell LLM's config by llm_id.

Get one voice's details by voice_id (path), e.g.

List all conversation flows in the account (current endpoint).

List all Retell LLMs in the account (current endpoint).

List all voices available to the account (built-in + any you've added/cloned).

Update a conversation flow.

Update a Retell LLM.

Launch a BATCH outbound calling campaign (one task per recipient).

Get the account's current and allowed call concurrency.

Run a batch of test-case definitions against an agent brain as a single job (AI QA).

Create a simulated-conversation TEST CASE for an agent (AI QA).

Delete a test-case definition.

Get a batch-test job's status and aggregate results by id.

Get one test-case definition by id.

Get one individual test run's result (path test_case_job_id) — the simulated conversation and per-metric pass/fail.

List batch-test jobs for an agent brain (current endpoint).

List test-case definitions for an agent brain (current endpoint).

List all individual test runs within a batch-test job (path test_case_batch_job_id) — current endpoint.

Register a call session for CUSTOM TELEPHONY / your own SIP trunk (advanced).

Re-run post-call analysis for a finished call (path call_id) — e.g.

Update a finished/stored call's metadata and a few settings.

Steer an IN-PROGRESS (ongoing) call in real time (path call_id) — override fields on the running call and/or push live agent control.

Update a test-case definition.

Create a new draft version of an agent (path agent_id) by branching from an existing one.

Start an outbound SMS chat.

Delete one specific version of an agent.

Permanently delete a chat and its messages.

List every saved version of a voice agent (path agent_id), so you can see history and pick a version to publish or update.

Re-run post-chat analysis for a finished chat (path chat_id), e.g.

Update a chat's metadata / stored settings.

Add sources (files / URLs / text) to an existing knowledge base.

Create a Knowledge Base (RAG) that agents can cite.

Delete a chat agent and all its versions.

Delete a whole knowledge base and all its sources.

Delete one source from a knowledge base.

List every saved version of a chat agent.

Update a chat agent.

Add a voice from a provider's community/marketplace library to your account so you can use it.

Clone a voice from sample audio.

Create a reusable Conversation Flow COMPONENT — a saved sub-graph of nodes you can drop into multiple flows.

Delete a reusable conversation-flow component.

Get a reusable conversation-flow component by id (path conversation_flow_component_id), including its node graph.

List all reusable conversation-flow components (current endpoint).

Search a voice provider's community/marketplace library by text.

Update a reusable conversation-flow component.

Run a single playground completion against an agent (path agent_id) — feed a message history and get the agent's next response without placing a real call.

List the tools an agent exposes through a connected MCP server.

List the account's data-export requests and their status/download links.