unpod logo
PlaygroundAboutBlogDocs
Github
Developer PlatformBuild, ship and observe voice agents on one platform.Platform
Build
IDE for Voice AgentsDesign and preview conversation flowsPlaybooksKick-start your agent with community playbooksFramework & SDKsPython, Node, REST and MCP
Ship
Contact CenterOpen-source contact center for large teamsTelephonyCarrier-grade telephony, numbers, HD qualitySpeech InfraSTT, TTS models, fault-tolerant orchestration
Observe
Observability & AnalyticsMonitor and debug every call
unpod logo

Realtime voice + messaging infrastructure. An open-source conversation framework and CPaaS, on metered telephony + speech - so any agent can call and text the real world.

GitHubDocumentation

Products

  • Developer Platform
  • Unpod (OSS CPaaS)
  • SuperDialog

Developers

  • Documentation
  • GitHub
  • API reference

Company

  • About
  • Privacy
  • Terms
© 2026 Unpod.ai · All rights reservedMade in India ❤️ · Built for the world

Build, Train & Ship conversational agents

An IDE to design conversations, observe live calls and ship phone-ready voice agents. Run them as a playbook, or as one LLM endpoint, with speech, telephony and observability built in.

Try Playground →Read the docs
BuildSuperDialog IDE→OptimizeRL Simulation→ShipInference Infra→ScaleOpen-source CPaaS→

Hover a stage to see how it works.

What teams ship on Unpod
AI receptionistLead qualificationAppointment bookingEMI & renewal remindersSupport triageOutbound follow-upCandidate screeningMultilingual supportOn-call pagerCoding-agent phoneAI receptionistLead qualificationAppointment bookingEMI & renewal remindersSupport triageOutbound follow-upCandidate screeningMultilingual supportOn-call pagerCoding-agent phoneAI receptionistLead qualificationAppointment bookingEMI & renewal remindersSupport triageOutbound follow-upCandidate screeningMultilingual supportOn-call pagerCoding-agent phone
TELEPHONY · SPEECH · AGENTIC INFRA

Take your voice agents to production, faster.

Unpod handles the phone and the speech. Text arrives at your playbook and you reply with text. The flow, the model and the customer data all stay on your side. Streaming STT, unified routing and neural TTS, end-to-end in under 330ms.

Try Playground →Read the docs
Your Live Stack
Playbook
Voice Profile
Build & Simulate
Deploy
Execute
Observe & Compliance
LAYER 01
Playbook
Design AI conversations with visual playbooks, prompts, tools, variables and knowledge to create production-ready agent workflows.
FlowsPromptsKnowledge
Live calls

Hear an agent handle a real call. Then ship your own on the same rails.

Realtime voice for appointments, reminders and support — on carrier-grade telephony and streaming speech. Tap any card to listen.

Live
AI Voice Agentunpod · call #4821
“Hello!”
2M+
minutes / mo
<330ms
end-to-end
24/7
uptime
Read the docs →
Live

AI receptionist

Answers inbound 24/7, qualifies intent, books meetings, warm-transfers with context.

Live

Phone-call your codebase

Call a number, speak a command — your coding agent runs it live over native MCP.

# pip install unpod
from unpod import Unpod

client = Unpod(api_key="up_...")
call = client.calls.create(
  agent_id="agt_123",
  webhook="https://you.dev/unpod"
)
✓ call placed · streaming to webhook

Runs in your stack

Python, Node, REST or MCP — all open source. Text in, text out, any LLM.

Proven in production

From first playbook to a live phone number.

0–4 days
to go live
<330ms
end-to-end latency
60% ↓
cost vs legacy stacks
2M+
minutes a month

Data residency

Pin storage to US, EU or India regions. Your choice, per project.

Encrypted end-to-end

TLS in transit, AES-256 at rest. Bring your own KMS keys.

Zero-retention audio

Streams aren’t stored by default. Opt in only if you need recordings.

Your model only

Transcripts train your own model when you ask. Never a shared or third-party one.

ISO
27001
ISO 27001
Information security
ISO
42001
ISO 42001
AI management
DPDP
India
DPDP India
Data protection
SOC 2
Type II
SOC 2 Type II
In progress
Explore Unpod

See Unpod in Action

Go live in weeks, not quarters. Unpod runs the call, reads the room, and knows exactly when to bring a human in.

SUPERDIALOG IDE

The Playbook Writes Itself

A copilot drafts your conversation flow, Talk Live previews it by voice, and every change is versioned. No wall-of-prompt text, no black-box tuning.

  • Copilot authors and refines the playbook alongside you
  • Talk Live to preview any flow the instant it's ready
N NovaTech Industries... DRAFT Editor Preview Docs Import Publish J Valid · 14 steps · journey: main DRAFT ✓ Saved just now Share Export Save 1 # Authored simple-playbook form. Compiles via superdialog.playbook.load_simple. 2 # Domain: Recruitment / HR - Exit Interview Feedback Agent 3 # Fictional employer - no real entity or rehire-guarantee implied. 4 name: "NovaTech Industries - Exit Interview Feedback Agent" 5 goal: >- 6 Conduct a structured exit interview call with a departing employee: 7 create a safe, non-judgmental space to share honest feedback, ask 8 structured questions about their reasons for leaving and overall 9 experience, capture themes for internal reporting, and close respectfully 10 regardless of the tone of feedback. NEVER argue with or dismiss critical 11 feedback, NEVER guarantee changes will be made based on their feedback, 12 and NEVER pressure them to withdraw their resignation or reconsider. 13 persona: 14 name: Ananya 15 gender: female 16 language: >- 17 Default English at the open; language confirmed once and LOCKED 18 (English or Hinglish). Hinglish = mix of English + simple Hindi - never 19 pure Hindi. 20 voice_style: >- 21 Warm, neutral, non-judgmental - this is the departing employee's 22 opportunity to be heard honestly. One question at a time. Never list 23 options with serial numbers. Pronounce only the first word of the 24 employee's name with 'ji' after it, or skip the suffix if contextually 25 more appropriate. 26 identity: >- 27 You are Ananya, conducting exit interviews for NovaTech Industries. You 28 create a safe space for honest feedback, ask structured questions, and 29 capture themes for internal reporting without arguing or being 30 defensive. Core principles: neutrality, genuine listening, respect for 31 the departure decision, confidentiality of individual responses where 32 possible. 33 opening: >- 34 Greet the departing employee, introduce yourself as Ananya from HR, thank 35 them for their time with NovaTech, and explain this call is an 36 opportunity to share honest feedback about their experience. 37 closing: >- 38 End politely, and only when the employee confirms no more questions or 39 the interview is complete. Use the appropriate closing variant (see 40 facts.closing_variants) and end the call. Agent building your playbook Talk Live Run optimize_playbook? ~2 min · 3 personas Approve Skip Test run 100% Completion 25% Data capture 90% Smoothness 0% Repairs 96% Empathy Voice LLM Pick one for the call claude-haiku-4-5 ~480 ms · est. gpt-4o-mini ~520 ms · est. gemma-4-31b-it ~600 ms · est. gpt-4.1-mini ~600 ms · est. Ask the agent to build or change something... + Claude Sonnet 4.6 ✦ Optimize Paste a website or doc link and the agent will read it.
SPEECH INFRA

Pick a Voice That Fits the Call

Streaming STT and TTS tuned for Indian languages and English. A curated set of voices per business type, previewable in one tap.

  • Multiple providers side-by-side: Unpod, ElevenLabs, Cartesia, Sarvam, Smallest, GNANI
  • Live latency and quality scores next to every voice, per language
Pick a Voice Profile Choose how your profile sounds. We've ordered these to fit your business type. Riya Female · UNPOD EN HI Quality Good Latency ~699ms Arjun Male · ELEVENLABS EN Quality Good Latency ~649ms Taniya Female · SARVAM HI Quality High Latency ~799ms Anika Female · CARTESIA MR Quality Good Latency ~700ms Shagun Female · CARTESIA EN Quality Good Latency — Shruti Female · GNANI HI Quality Good Latency ~700ms Zara Female · ELEVENLABS HI Quality Good Latency ~649ms Pooja Female · SMALLESTAI MULTI Quality Good Latency ~720ms Neha Female · CARTESIA EN HI Quality Good Latency ~799ms Not sure which one to pick? You can preview each voice to find the perfect match. 🎧 Need help choosing? ›
SHIP

Or Call It Like an API

Same playbook, published as one LLM endpoint the OpenAI SDK can already call — cURL, Python, JavaScript, or Go.

  • One endpoint, one model ID, drop-in with any OpenAI-style client
  • Stateful sessions via x_session_id, greet-first support built in
AI Identity Network Published AD STEP 4 OF 4 Deploy as endpoint Call it like any chat API cURL Python JavaScript Go curl https://pool.unpod.ai/v1/chat/completions \ -H "Authorization: Bearer $UNPOD_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"public:PB_7ZRM....", "messages": [{"role":"user", "content":"hi"}]}' from openai import OpenAI client = OpenAI(base_url="https://pool.unpod.ai/v1", api_key=os.environ["UNPOD_KEY"]) r = client.chat.completions.create( model="public:PB_7ZRM....", messages=[{"role":"user","content":"hi"}]) import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://pool.unpod.ai/v1", apiKey: process.env.UNPOD_KEY }); const r = await client.chat.completions.create({ model: "public:PB_7ZRM....", messages: [{role:"user",content:"hi"}] }); req, _ := http.NewRequest("POST", "https://pool.unpod.ai/v1/chat/completions", body) req.Header.Set("Authorization", "Bearer "+os.Getenv("UNPOD_KEY")) req.Header.Set("Content-Type", "application/json") resp, _ := http.DefaultClient.Do(req) Deploy as endpoint © 2024 AI Identity Network. All rights reserved. Privacy Policy Terms of Service
SHIP

One Number. Live Instantly.

The moment your playbook is ready, deploy it as a live voice agent with a real phone number — published and callable in seconds.

  • A working phone number the instant you publish. No porting, no waiting
  • Or point the OpenAI SDK at it in cURL, Python, JS or Go. One line change
AI Identity Network Published AD STEP 4 OF 4 Deploy as voice agent Get a phone number instantly YOUR NUMBER +91 92409 XXXXX Live the moment you deploy Deploy as voice agent © 2024 AI Identity Network. All rights reserved. Privacy Policy Terms of Service
AGENTS

Every Agent, Always Visible

Deployed agents, live sessions, and total minutes at a glance. Track dispatch health and drill into each worker's status in one view.

  • Live session counts and dispatch-error trends over 7 days
  • Per-agent cards showing active sessions, workers, and deploy age
Overview Telephony Overview Numbers Trunks Calls Speech Voice Profile Pipes Agents Workers Playbooks Sessions Playground Notifications Settings Unpod TVs Agent Workers ⓘ Docs + Deploy new agent ⌄ 🕑 Launch Console ↻ Auto-refresh off ⌄ AGENTS DEPLOYED ⓘ 3 ACTIVE SESSIONS ⓘ 0/200 AGENTS MINS ⓘ 25,965 mins Overview 📅 Past 7 days ⌄ AGENT SESSIONS SERVED ⓘ Total number of active sessions Agent dispatch errors 0 100 200 300 400 11 Jul 13 Jul 14 Jul 15 Jul 17 Jul 18 Jul 19 Jul Your agents Customer Support Agent ⋮ Active Sessions 32 Workers 3 RUNNING Deployed 6 days ago Sales Agent ⋮ Active Sessions 28 Workers 4 RUNNING Deployed 2 days ago Onboarding Agent ⋮ Active Sessions 19 Workers 2 RUNNING Deployed 1 day ago
OBSERVABILITY

Every Call, on One Console

Wallet balance, inbound vs outbound, missed calls, spend and success rate. The whole voice stack on one screen, updated live.

  • Real-time usage across numbers, agents and languages
  • At-a-glance stats: duration, missed calls, success rate, and spend
Overview Telephony Overview Numbers Trunks Calls Speech Voice Profile Pipes Agents Workers Playbooks Sessions Playground Notifications Settings Unpod TVs Unpod TVs Owner Start date → End date ⇩ Export Wallet Balance: 55.42 bits TOTAL 1,356 ▲ 92.3% CONNECTED 0 ▼ 0.00% Success Rate INBOUND CALLS 133 ▼ 5.7% OUTBOUND CALLS 1,223 ▲ 117.6% ACTIVE NUMBERS 3 TOTAL SPEND $0.58 ▲ 100.0% $ Usage Overview Daily ⌄ 0 90 180 270 360 Jun 19 Jun 24 Jun 29 Jul 03 Jul 08 Jul 11 Jul 14 Jul 18 Inbound vs Outbound Daily ⌄ 0 80 160 240 320 Inbound Outbound ✦ At a Glance Avg. Call Duration 01:42 Total Minutes 2,314 Missed Calls 12 Success Rate 98.6% Spam Calls Blocked 45
SDKs · open source

Runs in your stack.

Point the OpenAI SDK at your trained endpoint, or run the open-source framework in-house. Python, Node, REST or MCP. Text in, text out, any model.

~/ app.py
pip install openai
OpenAI SDK
SuperDialog
Node.js
REST
MCP
import os
from openai import OpenAI

# Already have an agent loop? Change one line.
client = OpenAI(
    base_url="https://api.unpod.dev/v1",
    api_key=os.environ["UNPOD_KEY"],
)

reply = client.chat.completions.create(
    model="agt_123",   # your trained playbook endpoint
    messages=[{"role": "user", "content": "Is Friday 4pm free?"}],
)
✓ 200 OK · 128ms
"Yes, Friday 4pm is free. Booked."
Read the docs →View on GitHub →
FAQ

Questions before you start.

The IDE, the models, integration and security: the things developers ask before shipping.

A conversation playbook: a versioned flow graph, not a wall of prompt text. A copilot writes and edits it with you, Talk Live previews it by voice, and Optimize runs it against simulated callers and tunes it. The playground is free and needs no login.
No. Playbooks run on any model through one-string model URIs: hosted APIs or your own vLLM / Ollama. Training a ~1B model on your data is the escalation for when a general model plateaus on your workflow, or the data can’t leave your side.
Yes. Point the playbook at any provider with a single URI, swap it at runtime, or register a custom provider. Self-hosted models get the full framework for free.
On a prompt-box platform your control ends at the prompt: a general model, hand-tuned, in a black box. Here the playbook is explicit and exportable, an RL environment tunes it in simulation, and when you need more you train a model you control.
The whole call: streaming STT and TTS tuned for Indian languages and English, numbers, PSTN, SIP and WebRTC across voice, SMS and WhatsApp. Your application sees one webhook, text in and text out, metered per minute.
No. The conversation framework and the contact-center CPaaS are open source and self-hostable. You pay only when live minutes run on Unpod’s metered infrastructure.
Data residency in US, EU and India. ISO 27001 and ISO 42001 certified, DPDP-compliant for India, SOC 2 Type II in progress. Audio is streamed, not stored by default; transcripts are encrypted at rest and train nothing but your own model.

Design the playbook. Train the model. Ship the number.

Start free in the playground, no login needed. Export the playbook to your own stack, or publish it live and pay only for metered minutes.

Try playground →View on GitHub ↗
✓ Free OSS framework✓ Self-host the CPaaS✓ Your data, your model✓ One endpoint change
Telephony, Speech & Agentic Infra for AI Agents - Unpod AI