DialSavvy MCP & APIComing

Your AI agents call through DialSavvy.

Agents in Claude, ChatGPT, Agentforce or your own code place calls, read outcomes and fetch transcripts, with the same checks as a rep's call.

Book a demo MCP comes with Command

The MCP server is coming to Command and Enterprise; tool names and scopes may change before launch. Webhooks are live on every plan; the REST API is coming, and the examples on this page are a preview.

Every request will get a rep's compliance checks before it rings.
AI calls and bulk adds will wait for a person on your team.
Agents will sign in with OAuth, scoped to one workspace.

Updated Oct 10, 2026

Agent requestsAcme · todayConcept1 call live

place_callSpeed-to-lead agent · Claude3:28 PM
Northwind TradersContact name hidden, number ending 43AI voice
Window openDNC checkedConsent checkedApproved by Jane Rivera
Ringing 0:04 · from +1 202 555 0119
add_to_campaignResearch agent · ChatGPT3:29 PM
Add 140 contactsto Trial sign-ups · OctoberNeeds a person
More than 100 contacts waits for a person on your team.
get_transcriptPipeline agent · Agentforce3:21 PM
Tidewater FreightContact name hidden, number ending 67Done
Read · ran on its own · returned as untrusted content

Same checks as a rep's call

Acme policydefaults

Read: On its own
AI-voice call: Needs a person
Over 100 contacts: Needs a person

Fig. 1 · Agent requests and approvals · Concept · fictional data

01 · Where it works

One server for the agent platforms you use.

Connect once, sign in and choose a workspace. Your agent will see only the tools its scopes allow, and every action it takes will be on the record.

OAuth sign-in, scoped to one workspaceNo API keys pasted into prompts. The consent screen shows the scopes and the workspace; disconnect a platform and its access ends.
An audit record for every actionEach one names the agent, the person who connected it and anyone who approved.
The same contract as the REST APIEvery tool maps one to one to a REST operation, so an agent can do what your own integration can, and nothing more.

ConnectionsAcme · Settings › API & MCPConcept

ClaudeRemote MCP connectorConnectedJane Rivera · calls:write calls:read
ChatGPTThe same server, as a ChatGPT appConnectedPriya Nair · calls:read lists:write
AgentforceMCP, plus voice handoff to Agentforce VoiceConnect
Your own agentsAny MCP client, or the REST API and webhooksActiveAPI key · Sam Okafor

DialSavvy MCP servermcp.dialsavvy.com/mcp · Streamable HTTP · OAuth 2.1

Fig. 2 · Connections in Settings · Concept · fictional data

02 · One request

One request. Three ways in.

Tools for your agent, endpoints for your code, signed webhooks back to your systems. The same checks run on every path.

mcp.dialsavvy.com/mcp

Speed-to-lead agentin Claude3 tool calls · 3:28 PM

list_calls3 found

{"outcome": "callback_requested", "since": "today"}

place_callRinging

{

"contact_id": "ct_0192b7e4",

"mode": "ai_voice",

"agent": "speed_to_lead",

"caller_id": "local_presence"

}

→ {"status": "approval_required", "approval_id": "apr_0192b7e6"}

// Jane Rivera approves in DialSavvy at 3:29 PM

→ {"status": "placed", "call_id": "call_0192b7f1"}

get_transcriptWhen the call ends

{"call_id": "call_0192b7f1"}

What DialSavvy did

Scope checkedcalls:write
Checked before it ringsWindow openDNC checkedConsent checked
Your policyAI-voice calls wait for a person.Approved by Jane Rivera · 3:29 PM
Placed on a local caller IDNorthwind TradersContact name hidden, number ending 43Ringing 0:04from +1 202 555 0119

Same checks as a rep's callOAuth 2.1 · scoped to Acme

Fig. 3 · An agent’s tool calls through MCP · Concept · fictional data

03 · Tools

What your agent can ask for.

Each tool is one REST operation. Your agent only sees the tools its scopes allow.

Calls

calls:write

place_callPlace a call with an AI voice, or queue it for a rep.
hand_to_repPass a live AI call or a contact to a rep's session.
set_outcomeSet or correct an outcome. The change is logged.

place_call · via a rep · 3:27 PMReason: “Asked about pricing on Tuesday.”Riverbend Credit UnionQueued for Jane Riverahand_to_rep · 3:31 PMAgent call with Northwind Traders handed to Leo Brandt, with a briefWarm transfer

Call results

calls:read

list_callsFind calls by outcome, rep, campaign or date.
get_transcriptThe transcript, wrapped as untrusted content for your agent.
get_summaryThe AI summary and suggested next step, labelled as AI.

AIget_summary · Northwind TradersWhy?

Wants pricing for a team of 12, starting in November. Asked for a callback Thursday at 10 AM.

Suggested next step: schedule the callback.

Lists and callbacks

lists:write

add_to_campaignAdd contacts to a rep's campaign. More than 100 at once waits for a person.
schedule_callbackBook a callback inside the contact's calling hours.
get_campaignProgress of one campaign: to call, retry later, done.

schedule_callback · 3:29 PMBlue Harbor CoThu 9:00 AM their timeInside calling hours

Checks and numbers

compliance:read

check_contactCalling window, do-not-call and consent for one contact.
list_numbersYour caller IDs and where each one is registered.
get_usageCredits used today and the balance left.

check_contact · 3:29 PMFernhill LogisticsDNC checkedOutside calling hours8:29 PM their time, opens 9 AM. The agent books a callback instead of calling.list_numbers · 3:29 PM+1 202 555 0119 · local presence, USCaller ID registered

Fig. 4 · Tool groups and what they return · Concept · fictional dataTool names and scopes may change before launch.

04 · Policy

Agents move fast. The rules don't bend.

Every call an agent asks for will get the same compliance checks as a rep's call. Your policy will decide what runs on its own and what waits for a person.

Every request takes the same path

Agent, API or rep: no shortcuts.

01The agent asksA tool call from Claude, ChatGPT or Agentforce, or a request to the REST API.place_call
02Scope checkedThe sign-in must carry the scope for this action, in this workspace.calls:write
03Compliance checksThe checks a rep's call gets, for the contact's country.Window openDNC checkedConsent checked
04Your policyIt runs on its own, or waits for a click from someone on your team.Approved by Jane Rivera
05Placed and loggedIt rings on a local caller ID. The audit log names the agent, who connected it and who approved.Local caller ID

When a check fails, the call never rings.

The agent gets the reason back as a standard problem response it can act on: outside calling hours, on a do-not-call list, or no consent on file. Nothing is dialed, and nothing is charged.

Fernhill LogisticsOutside calling hours

{

"title": "Outside calling hours",

"status": 422,

"their_time": "8:29 PM BST",

"opens_at": "2026-10-08T09:00:00+01:00"

}

Fig. 5 · The path every request takes · Concept · fictional data

You decide what waits for a person.

Owners and admins set the policy in Settings › API & MCP. New workspaces start with these defaults.

A person approves, inside DialSavvy.The approval is a click by someone on your team, in DialSavvy or on its card in the agent's chat. A model can't approve its own request.
Callers can't steer your agent.Transcripts reach your agent wrapped as untrusted content, so nothing said on a call becomes an instruction.
One switch stops every agent.An owner can refuse every agent request in the workspace at once, and turn access back on when ready.

Agent policy

Acme · Settings › API & MCP

Agents can actOn

ActionDefault

ReadLook up a call, an outcome or a transcript
Internal writeAdd a note, set an outcome, schedule a callback
Call through a repQueue a call; the next free rep places it
AI-voice callSpeed-to-lead callback, meeting confirmation. No consent on file, no AI call.A person always approves a first AI call
More than 100 contactsLoad a list into a rep's campaignBulk adds always wait for a person
Never through agentsRemove someone from do-not-call, buy numbers or credits, change plans, roles or caller IDs, delete recordings, edit rule packs, raise capsNot available to agents

Approvals go to owners and admins. Admins can name a different approver and set daily caps. The last row is fixed.

Fig. 6 · Agent policy in Settings · Concept · defaults shown

05 · Plans

Where it's included.

Calls, numbers and AI are paid from prepaid credits at published rates, whether a rep, an agent or your code starts them.

See pricing

Webhooks, and the REST API

Every plan

Dial
Power
Parallel
Command
Enterprise

Webhooks: live. Signed events to your endpoint for number and registration changes; call events are coming.

ComingREST API: read and write calls, outcomes, lists and numbers from your own systems, with API keys and public docs.

This call, started by an agentIllustrative5 min · US · recorded · AI notes

Calling, detection, recordingUS$0.104

Transcription and AI notesUS$0.047

You payUS$0.151

Example rates — the published rate card is final at launch. Same price as a rep's call.

MCP server

Coming

Command and Enterprise

Agents in Claude, ChatGPT or Agentforce get the tools on this page, under your policy. Enterprise is priced with you.

CommandEarly access

US$119per seat / month

Billed yearly · US$1,428 a year per seat, saves US$240

US$139 per seat / month billed monthly

Early-access pricing — final at launch

MCP questions

The MCP server is coming. Ask in your demo what it will do for your agents.

Book a demo

Can AI agents make phone calls through DialSavvy?

That is what the MCP server is for. It is coming to the Command plan and Enterprise, and will let an agent you build in Claude, ChatGPT, Agentforce or your own code place calls, read outcomes and fetch transcripts, with the same caller ID and compliance checks as a rep's call. Webhooks are live today; the REST API is coming.

What is an MCP server?

MCP (Model Context Protocol) is an open standard that lets AI assistants and agents use outside tools. An MCP server publishes those tools. DialSavvy's will publish tools for calls, call results, lists and callbacks, and checks and numbers.

Which plans include MCP?

The MCP server, which will let agents you build in Claude, ChatGPT, Agentforce and similar platforms place calls and use every DialSavvy capability, is coming to Command and Enterprise. Webhooks are live on every plan; the REST API is coming.

Next product

Next product: Telephony & caller ID

We are your carrier: numbers, registrations and local caller ID, set up for you before your first call.