LaunchedEditorial Listing

Cartesia Managed Agents

Cartesia · Cartesia Managed Agents: Hosted Voice Agents for Inbound and Outbound Phone Calls, Built on Cartesia's Own Speech Models

Open Cartesia Managed Agents

Cartesia Managed Agents is a hosted platform for building and running voice agents that answer and place phone calls, look up data and take actions in your systems mid-call, and run outbound batches of up to 5,000 calls. It suits developers and enterprise teams who want a configured agent (prompt, model, voice, tools) rather than their own voice stack, with published per-minute pricing that starts on a free plan.

PricingFreemium
Setupmedium
Runs onWeb · API
APIYes
Open sourceNo
DocsYes
Voice AIVoice AgentsTelephonyPhone CallsOutboundCustomer SupportSchedulingLead QualificationMultilingualDeveloper PlatformAPIWebhooksSOC 2HIPAAEnterpriseSpeech Synthesis

Best for

Developers and enterprise CX teams that want hosted inbound and outbound phone agents with low per-minute pricing, without assembling and tuning separate speech-to-text, LLM and text-to-speech providers

Not ideal for

Teams that need to bring their own LLM or key, place international calls without Twilio or a SIP carrier, run agents on their own infrastructure, or want a finished no-code support product with helpdesk and CRM connectors built in

Who it's for

Developers, product teams and contact center leaders building voice agents for customer support, scheduling and sales calls

Capabilities

  • Agents defined as configuration (instructions, greeting, model, voice, language, tools) in the Playground or through the REST API, with templates, Edit with AI, in-browser preview calls, publishing and version history
  • Managed LLM catalog with models from OpenAI, Anthropic and Google Gemini, chosen per agent with no provider account or API key, plus automatic or explicit fallback to up to three backup models
  • Speech handled by Cartesia's own models: Ink speech-to-text with noise suppression and keyterms for names and jargon, and Sonic voices with per-language voice presets
  • System tools for ending calls, transferring to phone numbers or SIP addresses, sending keypad (DTMF) tones, detecting voicemail and leaving a message, and pausing when a caller asks for a moment
  • Webhook tools that call HTTPS endpoints on your server mid-call and client tools that run functions in your app over the WebSocket API, up to 64 tools per agent
  • Knowledge bases built from folders of documents with metadata filters, uploaded in the Playground or by API and attached per agent (announced at launch; the migration guide still marks this as coming soon for managed agents)
  • Telephony through Cartesia-provisioned US numbers, numbers imported from Twilio, or your own carrier over SIP trunking with TLS signaling and optional SRTP
  • Outbound calling by single API request or batch calling of up to 5,000 recipients per batch, with per-recipient dynamic variables, a concurrency cap, retries and cancellation
  • Multilingual agents: a primary language plus additional languages the agent can switch to when the caller asks, with Cartesia citing 44 supported languages
  • Recordings and transcripts for every connected call, call webhooks for start, turn, end and summary events, and built-in plus custom LLM-as-a-judge metrics with export

Limitations

  • LLM token usage is billed separately on top of the per-minute agent rate, and Cartesia-provided numbers add $0.014 per minute, so the $0.03/min headline is not the full cost of a call
  • The $0.03/min rate is presented as a 50% launch discount from a previous $0.06/min, and Cartesia has not published when or whether the higher rate returns
  • Managed agents use only models in Cartesia's catalog; bringing your own LLM provider key is not supported on managed agents
  • Cartesia-provisioned numbers are US-only and can only call US destinations; international calling needs Twilio or your own SIP carrier
  • At launch, mid-call language switching covers English, Spanish, Hindi, French and Japanese, with the rest of the 44 languages described as coming soon
  • Cartesia stops hosting code-first Line SDK agents on December 1, 2026, so existing Line users must rebuild on Managed Agents or host their agent code themselves
  • A newly launched product: Cartesia's migration guide still lists knowledge bases as coming soon for managed agents, Line SDK agent-to-agent handoffs are not supported yet, and evaluations are free only for a limited time
  • Free and Pro plans are capped at 8 and 12 concurrent calls, and the Free plan includes only about 33 agent minutes a month and no commercial use license
  • You are responsible for TCPA and other calling-law compliance on outbound campaigns

Use cases

  • Answering an inbound customer support line, looking up orders or accounts through webhook tools and transferring to a human queue when needed
  • Booking, rescheduling and confirming appointments by phone against your own scheduling API
  • Calling new leads within minutes, qualifying them on your criteria and booking meetings with sales reps
  • Running an outbound batch of reminder, renewal or collections calls personalized per recipient with dynamic variables
  • Putting a voice agent behind an existing Twilio number or SIP carrier without changing phone providers
  • Embedding a talking agent in a web or mobile app over the WebSocket API

Our take

Cartesia Managed Agents is the voice agent platform to price out first if per-minute cost matters. At $0.03 a minute plus LLM tokens (and $0.014 on Cartesia numbers), it undercuts the published platform fees of Vapi and ElevenLabs Agents, and because Cartesia owns the speech recognition and voice models it can tune turn-taking and interruptions as one system, which Cartesia credits for its first-place showing on Artificial Analysis's blind Speech Agent Arena. What you give up is choice and maturity: you pick from Cartesia's LLM catalog rather than bringing your own, international calling needs Twilio or your own SIP carrier, and the platform is weeks old, with code-first Line SDK agents losing hosted support on December 1, 2026. It fits teams that want a hosted, configurable phone agent with good conversational feel at low cost; teams that need provider freedom should look at Vapi, and contact centers that want a finished platform with dialers, journeys and a human agent desktop should look at Regal.

Who should use it

Developers and enterprise CX, scheduling and sales teams that want hosted inbound and outbound phone agents with tool calling, published per-minute pricing and a free plan to start, and that are happy to choose models from Cartesia's catalog.

Who should skip it

Teams that must bring their own LLM key or run agents on their own infrastructure, businesses that need Cartesia-provided numbers outside the US, and non-technical small businesses looking for a ready-made receptionist with calendar and CRM connectors out of the box.

Strengths

  • Low published per-minute rate with a free plan and self-serve upgrades, rather than sales-only pricing
  • Speech recognition, voices and agent harness come from one vendor, and the agent ranked first on Artificial Analysis's blind Speech Agent Arena when checked on November 9, 2026
  • No LLM provider account to manage, with automatic model fallback
  • Flexible telephony: Cartesia numbers, imported Twilio numbers or your own SIP carrier
  • Outbound batches of up to 5,000 calls with per-recipient variables, plus inbound lines on the same agent
  • Built-in testing and call scoring, recordings, transcripts and webhooks; SOC 2 Type II, with HIPAA and GDPR listed and BAAs on Enterprise

Weaknesses

  • Headline price excludes LLM tokens and telephony, and the launch discount has no published end date
  • Closed model catalog with no bring-your-own LLM key on managed agents
  • Cartesia numbers are US-only for both origin and outbound destinations
  • Mid-call language switching covers five languages at launch
  • New product, with Line SDK hosting ending on December 1, 2026 and knowledge bases still marked coming soon in the migration guide

Cartesia Managed Agents pricing

Free

$0

Billed monthly

  • $1 of prepaid agent usage a month (about 33 minutes at $0.03/min)
  • 1 Cartesia phone number
  • 8 concurrent calls
  • 20K model credits for TTS and STT

Pro

$5/month

Billed monthly

  • $5 of prepaid agent usage a month
  • 3 Cartesia phone numbers
  • 12 concurrent calls
  • Commercial use license

Startup

$49/month

Billed monthly

  • $49 of prepaid agent usage a month
  • 5 Cartesia phone numbers
  • 20 concurrent calls

Scale

$299/month

Billed monthly

  • $299 of prepaid agent usage a month
  • 10 Cartesia phone numbers
  • 60 concurrent calls
  • Priority support

Enterprise

Custom

  • Volume pricing and custom concurrency
  • DPAs and BAAs
  • SSO and a shared Slack channel

Free tier limits: Free: $1 of agent usage a month (about 33 minutes), one Cartesia number, 8 concurrent calls and no commercial use license.

Note: Each plan's fee is credited back as prepaid voice agent dollars, and calls keep running past that balance at the plan's agent rate. Agent minutes cost $0.03, shown on the pricing page against a previous $0.06 and described in the launch post as 50%-off launch pricing; no end date is published. Telephony adds $0.014 per minute only on Cartesia-provided numbers (Twilio and SIP numbers are billed by your provider), and LLM usage during calls is billed separately at passthrough token rates. Evaluations are free for a limited time.

Technical specs

Modalities

Voice, Text

API pricing

Per call-minute: $0.03/min agent rate at launch pricing (previously $0.06/min), plus $0.014/min on Cartesia-provided numbers and LLM tokens billed separately

Available models

Managed LLM catalog with OpenAI, Anthropic and Google Gemini models (available models listed per account through the API)Cartesia Ink speech-to-textCartesia Sonic text-to-speech

Where Cartesia Managed Agents excels

Inbound support line with order lookups

Webhook tools let the agent fetch an order or account record from your API during the call, and the transfer system tool hands the caller to a human number or SIP address when needed.

Outbound appointment reminders

A batch of up to 5,000 recipients, each with its own name and appointment details as dynamic variables, dials in the background within a concurrency cap, with voicemail detection leaving a message when nobody answers.

Adding an AI agent to an existing phone system

Importing Twilio numbers or connecting a SIP trunk keeps current numbers and carrier rates, while Cartesia's telephony fee applies only to Cartesia-provided numbers.

Comparing models before going live

Switching the LLM creates a new agent version, and custom LLM-as-a-judge metrics plus recordings and transcripts let you score test and production calls before and after the change.

Cartesia Managed Agents vs. competitors

Cartesia Managed Agents vs. Vapi

Vapi is a provider-agnostic orchestration platform: you pick the speech-to-text, LLM, text-to-speech and telephony providers and pay their costs on top of $0.05/min hosting. Cartesia Managed Agents bundles its own speech models with a managed LLM catalog at $0.03/min plus LLM tokens, trading provider choice for one tuned system and a lower platform fee.

Cartesia Managed Agents vs. ElevenLabs Agents

ElevenLabs Agents is the closest match: a speech vendor's own agent platform with knowledge bases, tools and telephony. ElevenLabs charges $0.08/min with LLM usage billed separately and no telephony fee (you connect Twilio or a SIP trunk), supports bring-your-own LLMs and more channels (WhatsApp, SMS, email and chat), and lists 90+ languages. Cartesia is cheaper per minute at launch pricing but limits you to its model catalog and phone and WebSocket channels.

Cartesia Managed Agents vs. Regal

Regal is a sales-led enterprise contact center platform that wraps voice, SMS and chat agents with dialers, journeys, CRM connectors and a human agent desktop, at custom prices. Cartesia Managed Agents is a self-serve, published-price platform focused on the voice agent itself, leaving CRM and contact center plumbing to your own webhooks.

Cartesia Managed Agents vs. Famulor

Famulor is a no-code phone agent platform for small and mid-sized businesses with booking, CRM and WhatsApp follow-up built in. Cartesia Managed Agents is more developer-oriented, connecting to your systems through webhook and client tools, with SIP and Twilio telephony and outbound batches aimed at larger call volumes.

Frequently asked questions

What is Cartesia Managed Agents?

It is Cartesia's hosted platform for voice agents, launched on October 29, 2026. You configure an agent's instructions, LLM, voice, languages and tools in the Playground or by API, connect a phone number, and Cartesia runs listening, turn-taking, interruptions and tool calls on its own speech models. Agents answer inbound calls, place outbound calls and batches, and can also be reached from apps over WebSocket.

How much does Cartesia Managed Agents cost?

Agent minutes cost $0.03, which Cartesia calls 50%-off launch pricing and shows against a previous $0.06 rate, with no end date published. Calls on a Cartesia-provided number add $0.014 per minute, and LLM usage is billed separately. Plans are Free ($1 of agent usage a month, about 33 minutes), Pro ($5/month), Startup ($49/month) and Scale ($299/month), each crediting its fee as prepaid agent usage, plus custom Enterprise.

Which LLMs can a Cartesia agent use? Can I bring my own key?

You pick a model per agent from Cartesia's managed catalog, which includes OpenAI, Anthropic and Google Gemini models, and Cartesia bills token usage as passthrough. You do not need, and cannot plug in, your own provider key on managed agents. Teams that need full control can still run a code-first agent built with the open-source Line SDK on their own servers and point Cartesia at it.

Can I use my own phone numbers?

Yes. Besides Cartesia-provisioned US numbers, you can import numbers from Twilio or connect your own carrier through SIP trunking. Cartesia numbers are US-only and call only US destinations; international calls go through Twilio at your Twilio cost or through your carrier.

Does it support outbound calling campaigns?

Yes. A single API request places one call, and batch calling queues up to 5,000 recipients per batch, each with its own dynamic variables, dialing in the background up to a concurrency limit you set. You are responsible for TCPA and other calling-law compliance.

What languages does it support?

Cartesia says agents can be built in 44 languages. You set a primary language and can add others the agent switches to when the caller asks; at launch, mid-conversation switching covers English, Spanish, Hindi, French and Japanese, with all 44 described as coming soon.

Is Cartesia HIPAA compliant?

Cartesia states it is SOC 2 Type II certified, and its Managed Agents page lists HIPAA and GDPR. DPAs and BAAs are part of the Enterprise plan.

Cartesia Managed Agents vs Vapi: what is the difference?

Vapi is an orchestration layer where you choose and pay separately for speech-to-text, LLM, text-to-speech and telephony providers, on top of a $0.05/min hosting fee. Cartesia runs its own speech models and a managed LLM catalog for $0.03/min plus LLM tokens, which is simpler and cheaper on the platform fee but gives you less choice of providers.

Integrations & fit

TwilioSIP trunkingOpenAIAnthropicGoogle GeminiWebSocket APIREST APIWebhooksPython SDKJavaScript/TypeScript SDK
Good fit forStartup / small team, Enterprise
Pricing modelFreemium· Free tier available
See pricing on Cartesia Managed Agents →

Alternatives to consider

About Cartesia Managed Agents

Cartesia, the company behind the Sonic text-to-speech and Ink speech-to-text models, launched Managed Agents on October 29, 2026. You define an agent as configuration rather than code: a system prompt and greeting, an LLM from Cartesia's managed catalog (OpenAI, Anthropic and Google Gemini models, with automatic fallback, and no provider API key needed), a voice, a primary language plus optional additional languages, and tools. You can do that in the Playground, which offers templates, an Edit with AI button, in-browser test calls and versioned publishing, or through the REST API. Tools come in three kinds: built-in system tools (end call, transfer to a phone number or SIP address, send keypad tones, voicemail detection, skip turn), webhook tools that call HTTPS endpoints on your server, and client tools that run in your app over the WebSocket API. Cartesia's launch post also announces knowledge bases built from folders of your documents, although its Line SDK migration guide still lists attaching knowledge bases to managed agents as coming soon. For telephony you can use a Cartesia US number, import numbers from Twilio, or bring your own carrier over SIP, and the same agents can be reached from web and mobile apps over WebSocket. Outbound work runs as single API calls or batches of up to 5,000 recipients, each personalized with dynamic variables. Every connected call keeps a recording and transcript, call webhooks stream start, turn, end and summary events, and you can score calls with built-in metrics and your own LLM-as-a-judge metrics. Cartesia's pitch is that its speech models and agent harness were designed together, which it says improves turn-taking and interruptions; on Artificial Analysis's Speech Agent Arena, a blind human-preference leaderboard, Managed Agents was ranked first when checked on November 9, 2026 (with a confidence range overlapping the second-placed Gemini model), and Cartesia reports the top accuracy and experience scores on ServiceNow's EVA-Bench. Pricing is the clearest differentiator against Vapi and ElevenLabs Agents: agent minutes cost $0.03, which the launch post calls 50%-off launch pricing and the pricing page shows against a previous $0.06 (no end date is published), Cartesia numbers add $0.014 a minute, and LLM tokens are billed separately at passthrough rates. The tradeoffs are a closed model catalog (no bring-your-own LLM key on managed agents), US-only Cartesia numbers, and a young product: Cartesia stops hosting code-first agents built with its earlier Line SDK on December 1, 2026, so teams on that SDK must migrate or self-host.

Updates from Cartesia Managed Agents

LaunchCartesia launches Managed Agents

Cartesia launched Managed Agents, a hosted voice agent platform with telephony, tool calling, knowledge bases and outbound batch campaigns, built on its own speech models. Agent minutes launched at $0.03, half the previous $0.06 rate, with LLM usage billed separately.

Are you the founder? Claim this listing →