Connect AI tools with MCP

Ody has a built-in MCP (Model Context Protocol) server at https://api.ody.co/v1/mcp — add it to ChatGPT, Claude, Cursor, or any MCP client with an API key, and the assistant can manage contacts, work your conversations, search messages, send texts, buy phone numbers, register 10DLC texting, place outbound AI calls, and configure your AI agent in your workspace.

Before you start

• You need an API key — see Get your Ody API key. For an assistant that should only read, create the key with just the read scopes. To experiment without sending real texts, use an ody_test_… key — see Build safely with test mode.
• MCP is included on every plan, starting with Starter.

Step-by-step

1. Click Settings in the sidebar, then Developer API.

2. Scroll to the Connect to ChatGPT & Claude card. It shows the MCP endpoint — https://api.ody.co/v1/mcp — with a Copy button, and a ready-to-paste client config.

3. In your AI tool's MCP/connector settings, add a server using this config (with your real key):

{
  "mcpServers": {
    "ody": {
      "url": "https://api.ody.co/v1/mcp",
      "headers": { "Authorization": "Bearer ody_live_…" }
    }
  }
}

4. Ask the assistant something like "give me a summary of my Ody workspace" to confirm the connection.

What the assistant can do

The MCP server exposes twenty tools, each gated by the same scope its REST endpoint uses:

Tool	What it does	Required scope
list_contacts	List or search your contacts	contacts:read
create_contact	Add a contact	contacts:write
update_contact	Update a contact's name, company, phone, or email	contacts:write
delete_contact	Permanently delete a contact	contacts:write
list_conversations	List open/done conversations	conversations:read
get_conversation	Read a full thread	conversations:read
set_conversation_status	Mark a conversation open or done	conversations:write
search_messages	Search message bodies, voicemail transcripts, and call notes	conversations:read
workspace_summary	Snapshot: contact count, open/done counts, unread total, 7-day volume	conversations:read
wait_for_message	Wait (up to 25 s) for a new inbound message in a conversation	conversations:read
send_message	Send an SMS (to a conversation or a phone number)	messages:send
list_numbers	List your workspace numbers	numbers:read
search_numbers	Search phone numbers you can buy	numbers:manage
buy_number	Buy a number — billed like an in-app purchase	numbers:manage
get_messaging_status	Check 10DLC texting registration status	numbers:manage
submit_10dlc_registration	File your 10DLC business + campaign (billed)	numbers:manage
place_call	Have your published AI agent place an outbound call	calls:place
get_agent	Read your AI agent's config and credit usage	agent:manage
configure_agent	Update the agent's config; pass publish: true to go live	agent:manage
connect_agent_to_number	Route one of your numbers to the AI agent	agent:manage

If the key is missing a scope, that tool simply isn't offered to the assistant — so a read-only key produces a genuinely read-only assistant. Everything runs with the permissions of the member who created the key, in that workspace only.

Two things worth knowing:

• wait_for_message lets an assistant hold a live SMS conversation: it sends a text with send_message, waits up to 25 seconds for the reply, and responds — calling the tool again if the wait times out.
• The number, 10DLC, call, and agent tools are the same "Agent Phone" surface as the REST API — same billing, same rules, and you can take over in the app at any moment. See Buy numbers and register texting via the API and Place AI calls and manage Astra via the API.

With a test key, send_message records a simulated message, buy_number and place_call simulate (nothing bought or dialed), agent publishing is simulated, and submit_10dlc_registration files in mock mode — no fee, no registry submission.

Related articles

• Get your Ody API key
• Build safely with test mode
• API overview and authentication