Telenow — AI Phone Calls & Voice Agents

What Telenow — AI Phone Calls & Voice Agents does

Telenow — AI Phone Calls & Voice Agents exposes Telenow’s phone and voice-agent functions through a remote MCP endpoint. An MCP client can use selected Telenow numbers to place real calls, inspect inbound and outbound call history, and retrieve a call’s status, outcome, summary, and transcript.

The server also supports voice-agent management. Clients can list existing agents and available voices, create a single-context agent from a system prompt, or create a multi-step agent represented as a graph. Agents created by the assistant can be updated later. The estimate_cost tool provides a per-minute USD estimate for an agent or planned configuration.

Campaign support covers drafting, targeting, starting, pausing, and reviewing outbound campaigns. Creating a campaign only produces a draft; dialing begins only when the campaign is started.

How it works

The hosted endpoint is https://api.telenow.ai/mcp. It uses Streamable HTTP with stateless JSON responses and supports several MCP protocol versions. Authentication uses OAuth 2.1 with PKCE and dynamic client registration, or an organization API key sent as a Bearer authorization header.

When connecting, the user signs in and selects the organization, phone numbers, and permissions available to the assistant. The assistant can call only from the selected numbers. It must request confirmation before an action reaches real people, including an individual call or campaign activity.

Permissions are represented by scopes such as calls:read, calls:write, agents:write, agents:tools, and campaigns:write. Agent tools, webhooks, and transfers remain disabled unless the agents:tools scope is granted. A read-only API key exposes read-only tools, while a write-capable key provides the organization’s full permitted access.

Setup and configuration

Telenow — AI Phone Calls & Voice Agents requires a Telenow account with at least one phone number. There is no local package or process to install because Telenow hosts the service.

Add the remote MCP URL in a compatible client. Claude, Cursor, VS Code, Windsurf, and Cline can connect using the URL and OAuth where supported. Claude Code can register the endpoint with its HTTP transport, and Gemini CLI provides an extension-install command for the repository. Clients without OAuth can send an organization API key in the request headers:

{
  "url": "https://api.telenow.ai/mcp",
  "headers": { "Authorization": "Bearer vai_live_…" }
}

The repository contains client configuration and documentation; the hosted endpoint is the component that handles requests.

Tools and capabilities

The 17 documented tools are grouped into three areas:

• Calls: list_numbers, place_call, get_call, and list_calls.
• Agents: list_agents, get_agent, create_agent, create_flow_agent, update_agent, list_voices, and estimate_cost.
• Campaigns: list_campaigns, get_campaign, create_campaign, add_campaign_targets, start_campaign, and pause_campaign.

add_campaign_targets accepts up to 1,000 people. Pausing a campaign allows calls already in progress to finish. Call and campaign charges come from the organization’s Telenow wallet and are subject to its spending limits. A typical setup is documented at about $0.049 per minute; use estimate_cost for an exact quote.

Limitations and notes

Telenow — AI Phone Calls & Voice Agents can act only within the selected organization, numbers, and permission scopes. It cannot change agents or campaigns that it did not create. The service also applies a rate limit of 3,000 requests per 15 minutes per connection or API key.

The assistant does not see or configure the user’s API keys. Calls consume wallet balance, and costs may vary by the selected voice, model, transcription provider, carrier, or other configuration. Components using the customer’s own vendor keys or carrier are not billed by Telenow. The repository is MIT licensed, while the hosted Telenow service is proprietary and governed by Telenow’s terms.