Glossary/87 terms
AI phone glossary
01
Phone network · Numbers
- 070 numberKorean VoIP number
- A 070 number is the Korean phone number range assigned to internet-phone (VoIP) services; it is not tied to an area code, so you can make and receive calls with the same number anywhere. KAT Cloud issues one 070 number when you sign up, and on a paid plan you can add more by API or MCP.Read more on our blog (in Korean)
- Representative number15xx · 16xx · 18xx
- A representative number is a nationwide single number such as 15xx, 16xx or 18xx that is used without an area code, typically to route calls to one number across multiple lines or branches. A KAT Cloud representative number costs KRW 30,000/month per number (VAT excluded).Read more on our blog (in Korean)
- Public switched telephone networkPSTN
- The public switched telephone network (PSTN) is the general telephone network operated and interconnected by fixed-line and mobile carriers. For an AI agent to talk to mobile phones or landlines, it needs a number and lines connected to this network.Read more on our blog (in Korean)
- PBXPrivate Branch Exchange
- A PBX (private branch exchange) is switching equipment or software that connects internal extensions within a company and links them to external lines. Connecting it to KAT Cloud over a SIP trunk lets you add AI call handling to an existing PBX.Read more on our blog (in Korean)
- Concurrent call linesConcurrent calls
- Concurrent call lines are the number of calls that can be in progress at the same time. KAT Cloud includes 10 lines on every plan and can scale up to 10,000 lines at KRW 100,000/month per 100 lines.Read more on our blog (in Korean)
- Erlang B
- Erlang B is a formula that calculates the probability that a call is blocked (grade of service) from traffic volume (in erlangs) and the number of lines; it is used to estimate how many concurrent call lines you need.Read more on our blog (in Korean)
- Facilities-based telecommunications carrierFacilities-based carrier
- A facilities-based telecommunications carrier is a business registered under Korea's Telecommunications Business Act that provides basic telecommunications services such as telephone and internet. Korea AI Telecom Inc., the operator of KAT Cloud, is a facilities-based telecommunications carrier (registration No. 1-01-25-0027).Read more on our blog (in Korean)
- Special category value-added telecommunications provider
- A special category value-added telecommunications provider is a business that provides value-added telecommunications services, such as text message sending, for which Korea's Telecommunications Business Act requires separate registration. Korea AI Telecom Inc. (KAT), the operator of KAT Cloud, is registered for its text messaging business (No. 3-01-25-0010).Read more on our blog (in Korean)
- Caller ID pre-registration
- Caller ID pre-registration is a Korean rule that allows text messages to be sent only from numbers that have been registered and verified in advance, to prevent number theft and caller ID spoofing.Read more on our blog (in Korean)
- DIDDirect Inward Dialing
- A DID is an individual phone number assigned so that outside callers can reach a specific extension or service directly without going through an operator or switchboard menu.If you bundle several DIDs on one SIP trunk, you can route calls to staff extensions or to AI agents depending on which number was dialed. In cloud telephony, each number is often treated as a DID with its own routing rules.Related termsSIP trunkPBX070 numberRead more on our blog (in Korean)
- Caller IDCLI · Calling Line Identification
- Caller ID (CLI) is the feature that shows the caller's phone number on the recipient's screen, and the number information itself.In SIP, the From header and the P-Asserted-Identity header determine the displayed number. In Korea you may display only numbers that have been assigned to you or that you are authorized to use; displaying a false number is prohibited as caller ID spoofing.Related termsP-Asserted-IdentityCaller ID spoofingCaller ID pre-registrationRead more on our blog (in Korean)
- Call forwarding
- Call forwarding is a feature that automatically passes calls made to your number on to another number you have set in advance.It is often split by condition: always forward, forward when busy, or forward when unanswered. KAT Cloud provides a forwarding number feature that passes incoming calls to a number you specify.Related termsCall transferDiversion headerHandoffRead more on our blog (in Korean)
- Call transfer
- Call transfer is the action of moving a call that is already connected to another person or number.There are blind transfers, which pass the caller straight through, and attended transfers, where you first speak with the person taking the call and then connect them. In SIP, transfers are requested with the REFER method (RFC 3515). A handoff, where a call handled by AI is passed to a staff member, is also a form of call transfer.Related termsHandoffCall forwardingPBXRead more on our blog (in Korean)
- Number portability
- Number portability is the arrangement that lets you keep your existing phone number when you switch telecommunications carriers.Not only mobile numbers but also landline and internet-phone (070) numbers can be ported between services of the same type. Identity verification and cancellation with the previous carrier are usually handled together, and incoming calls may be briefly unstable during the switch. KAT Cloud also supports number porting.Related termsCertificate of service use070 numberRepresentative numberRead more on our blog (in Korean)
- Certificate of service useProof of service
- A certificate of service use is a document issued by a telecommunications carrier confirming who uses a given phone number and for which service.It is often used to prove ownership of and the right to use a number when pre-registering a caller ID for text messages or applying for number porting. It is usually issued through your carrier's customer center or website, and if the account holder and the applicant are different, a power of attorney may also be required.Related termsCaller ID pre-registrationNumber portabilityCaller IDRead more on our blog (in Korean)
- Call unitsDosu
- Call units (dosu) are a billing unit that counts phone usage by converting call time or call count according to a fixed standard.Local calls in Korea used to be billed by charging one unit for every fixed period, and the term "units used" still appears in settlements between carriers and in communication records. Cloud phone services usually charge per second or per minute instead of by units.Related termsInterconnectionCommunication confirmation data
- Interconnection
- Interconnection is the linking of different carriers' telephone networks so that a subscriber of one carrier can call a subscriber of another.A call from a 070 number to a mobile phone also passes through interconnection between the internet-phone carrier and the mobile carrier. Carriers settle interconnection fees for these links, and that cost is reflected in outbound call rates.Related termsPublic switched telephone networkFacilities-based telecommunications carrierCall unitsRead more on our blog (in Korean)
- Value-added telecommunications providerValue-added service provider
- A value-added telecommunications provider is a business that provides telecommunications services other than basic telecommunications services, operating after filing a report under Korea's Telecommunications Business Act.Many lease lines from facilities-based carriers and provide services on top of them. Special categories defined separately by law, such as bulk text message sending, require registration rather than a report.Related termsFacilities-based telecommunications carrierSpecial category value-added telecommunications providerRead more on our blog (in Korean)
- IVR · ARSInteractive Voice Response
- IVR (commonly called ARS in Korea) is an automated answering system that sorts and handles calls with recorded prompts and keypad input.It announces menus such as "Press 1 for reservations, 2 for support" and takes DTMF input to route callers to the right department or information. A voice AI agent differs in that, instead of making callers choose from a menu, it understands what they say and handles or routes the request directly.Related termsDTMFACDVoice AI agentRead more on our blog (in Korean)
- ACDAutomatic Call Distribution
- ACD is a contact center feature that automatically distributes incoming calls to available agents according to set rules.Common rules include round robin, longest-idle agent first, and skill-based routing. When all agents are busy, callers are placed in a queue and hear announcements, and some setups add an overflow rule that passes calls to AI when the wait gets long.Related termsIVR · ARSCTIConcurrent call linesRead more on our blog (in Korean)
- CTIComputer Telephony Integration
- CTI is technology that integrates the phone system with business computer systems so that call information and customer information can be handled together.Typical examples are showing customer details on the agent's screen when a call comes in, or placing a call by clicking a button in the CRM. With cloud telephony that offers APIs and webhooks, the same is done by receiving call events and recording them in the CRM.Related termsWebhookACDPBXRead more on our blog (in Korean)
02
Protocols · Media
- SIPSession Initiation Protocol
- SIP is the standard protocol (RFC 3261) used in internet telephony to exchange the signals that start, modify and end calls. A call is requested with INVITE, answered with 200 OK and ended with BYE.Read more on our blog (in Korean)
- SIP trunk
- A SIP trunk is a virtual line that connects a PBX or voice server to a carrier's telephone network over SIP. Direct SIP connection to KAT Cloud is free on every plan.Read more on our blog (in Korean)
- SDPSession Description Protocol
- SDP is the format that describes which codecs, addresses and ports will be used to exchange audio when a call is set up; it is carried in the body of SIP messages so both sides can agree on media parameters (offer/answer).Read more on our blog (in Korean)
- RTPReal-time Transport Protocol
- RTP is a protocol (RFC 3550) that splits real-time media such as call audio into small packets tagged with sequence numbers and timestamps. Once SIP sets up a call, the actual audio flows over RTP.Read more on our blog (in Korean)
- SRTP · SIP over TLS
- SRTP encrypts RTP audio, and SIP over TLS encrypts SIP signaling. Used together, both the call signaling and the audio are encrypted.Read more on our blog (in Korean)
- G.711PCMU / PCMA
- G.711 is the baseline codec of the telephone network, sampling audio at 8 kHz and sending it at 64 kbps; it comes in two variants, μ-law (PCMU) and A-law (PCMA).Read more on our blog (in Korean)
- Opus
- Opus is an open voice and audio codec with low latency and variable bitrate; it supports bandwidths from narrowband to fullband and is widely used in WebRTC and internet voice calls.Read more on our blog (in Korean)
- DTMFDual-Tone Multi-Frequency
- DTMF is the signal made of two combined frequencies that is produced when you press a phone keypad, used for menu selection and number entry. In internet telephony it is usually sent as RTP events using the RFC 4733 method.Read more on our blog (in Korean)
- Jitter buffer
- A jitter buffer briefly collects voice packets that arrive over the network at irregular intervals and plays them out at a steady pace to reduce choppiness. A larger buffer is more stable but adds delay.Read more on our blog (in Korean)
- PCMU · μ-lawG.711 μ-law
- PCMU is the μ-law variant of the G.711 codec, a format that samples audio at 8 kHz, applies 8-bit logarithmic compression and sends it at 64 kbps.It is mainly used in North American and Japanese telephone networks and uses static payload type 0 in RTP. KAT Cloud SIP connections currently use G.711 μ-law (PCMU) by default, so PCMU must be in the codec list of the PBX or voice server you connect.Related termsPCMA · A-lawG.711SDPRead more on our blog (in Korean)
- PCMA · A-lawG.711 A-law
- PCMA is the A-law variant of the G.711 codec, a format with the same 64 kbps as μ-law but a different compression curve.It is used in telephone networks in Europe and many other regions, and its static RTP payload type is 8. If the two sides use different codecs, audio must be converted (transcoded) in between, which can affect quality and latency. KAT Cloud supports A-law (PCMA) along with μ-law (PCMU).Related termsPCMU · μ-lawG.711RTPRead more on our blog (in Korean)
- RFC 4733telephone-event
- RFC 4733 is the standard that defines how to send DTMF key presses as separate RTP event packets (telephone-event) rather than as audio.It replaces the older RFC 2833, and both sides agree on support by listing the telephone-event payload in SDP. Its advantage is that digits are delivered accurately even when key tones in the audio are distorted by compressed codecs or packet loss.Related termsDTMFRTPSDPRead more on our blog (in Korean)
- ptimePacketization time
- ptime is the value that sets how many milliseconds of audio go into one RTP packet; it is written in SDP so both sides agree.20 ms is the most common value in telephone networks, and a 20 ms G.711 packet carries 160 bytes of audio data. A larger value reduces the number of packets and header overhead, but increases delay and makes dropouts worse when packets are lost.Related termsRTPSDPJitter bufferRead more on our blog (in Korean)
- RTCPRTP Control Protocol
- RTCP is a control protocol exchanged alongside the RTP audio stream that reports quality statistics such as packet loss rate, jitter and round-trip delay.It is defined in the same RFC 3550 as RTP and usually uses the port number right after the RTP port. When diagnosing call quality problems, the loss rate and jitter values in RTCP reports are the basic data.Related termsRTPMOSPacket lossRead more on our blog (in Korean)
- INVITESIP INVITE
- INVITE is the SIP request method sent to the other party to start a call.Its body carries SDP proposing the codecs and audio address to use, and the other side responds with 100 Trying, 180 Ringing and 200 OK in that order. Once the requester sends ACK, the call is established and audio flows over RTP.Related termsSIPSDPSIP response codesRead more on our blog (in Korean)
- REGISTERSIP REGISTER
- REGISTER is the request a SIP device or PBX sends to register its current address with a server so it can receive incoming calls.The server usually demands authentication with a 401 or 407 response, and the device registers again with a digest value computed from its username and password. Registrations expire, so they must be renewed periodically. SIP trunks sometimes use IP authentication, identifying the peer by IP address, instead of registration.Related termsSIPSIP trunkSIP response codesRead more on our blog (in Korean)
- SIP response codes
- SIP response codes are three-digit numbers that report the result of a request; by the first digit they are grouped as 1xx provisional, 2xx success, 3xx redirection, 4xx client error, 5xx server error and 6xx global failure.During integration you will often see 401/407 (authentication required), 403 (forbidden), 404 (number not found), 486 (busy), 488 (not acceptable, e.g. codec mismatch) and 503 (service unavailable). Checking the Reason header that comes with the response code helps narrow down the cause.Related termsSIPINVITEREGISTERRead more on our blog (in Korean)
- Diversion headerDiversion · History-Info
- The Diversion header tells you, on a forwarded SIP call, which number the call was originally made to before it was forwarded.It is described in RFC 5806, and the more general History-Info header (RFC 7044) is also used for the same purpose. When several numbers are forwarded to one AI server, knowing the original number lets you apply a different greeting or task per number. Whether the header is passed through depends on the carriers and equipment settings along the path.Related termsCall forwardingSIPDIDRead more on our blog (in Korean)
- P-Asserted-IdentityPAI
- P-Asserted-Identity is a header that conveys the verified identity (number) of the caller within a trusted SIP network (RFC 3325).A device can write anything in the From header, but PAI is inserted and trusted only within segments that have a trust relationship, such as between carriers. It is used as the basis for determining the actual caller in caller ID display and spoofing prevention.Related termsCaller IDCaller ID spoofingSIPRead more on our blog (in Korean)
- SBCSession Border Controller
- An SBC is equipment or software placed at the border between different SIP networks that relays signaling and audio and handles security, address translation and codec conversion.Placed between a corporate PBX and a carrier's SIP trunk, it hides private IPs, filters abnormal requests and reconciles differences between SIP dialects on each side. SBCs often solve the one-way audio problem that occurs when a PBX behind NAT is connected to an external trunk.Related termsNAT traversalSIP trunkPBXRead more on our blog (in Korean)
- NAT traversal
- NAT traversal refers to techniques that let equipment inside a private network exchange SIP signaling and RTP audio correctly even though it passes through NAT, which translates addresses to a public IP.SIP and SDP bodies contain IP addresses and ports, and if equipment behind NAT writes its private IP, the other side sends audio to the wrong place and only one side can be heard. This is solved with methods such as external address settings, rport, STUN/TURN/ICE and SBCs.Related termsSBCRTPSDPRead more on our blog (in Korean)
- MOSMean Opinion Score
- MOS is a metric that expresses call audio quality on a scale from 1 (bad) to 5 (excellent).Originally it was the average of scores given by human listeners, but in practice values estimated from packet loss, jitter and delay are widely used. Poor audio quality lowers not only what people hear but also speech recognition accuracy, so it is an important operating metric for AI phone calls too.Related termsPacket lossJitter bufferRTCPRead more on our blog (in Korean)
- Packet loss
- Packet loss is when some voice RTP packets on the network never reach the other side.Losing one 20 ms G.711 packet drops 20 ms of audio, and once loss exceeds just a few percent you can hear choppiness and garbled speech. Causes include congested links, wireless segments and firewall processing delays, and you can check the loss rate in RTCP statistics.Related termsMOSRTCPJitter bufferRead more on our blog (in Korean)
- WebSocketRFC 6455
- WebSocket is a protocol that, after a connection is opened over HTTP, lets a server and client keep exchanging data in both directions over a single TCP connection (RFC 6455).In voice AI integrations it is often used as the channel for exchanging chunks of call audio in real time. KAT Cloud provides a bidirectional voice stream over WebSocket (G.711).Related termsVoice streamWebRTCSIP trunkRead more on our blog (in Korean)
- WebRTCWeb Real-Time Communication
- WebRTC is a set of standard technologies that let browsers and apps exchange real-time audio and video without plugins.It finds connection paths with ICE, encrypts audio with DTLS-SRTP, and commonly uses the Opus codec. To connect to the regular telephone network, it has to go through a SIP gateway and be converted to a telephone-network codec such as G.711.Related termsOpusSRTP · SIP over TLSNAT traversalRead more on our blog (in Korean)
- Voice streamMedia stream
- A voice stream is an integration method that cuts the audio exchanged during a call into real-time chunks, sends them to an external server, and plays the audio produced by that server back into the call.Its advantage is that you can attach your own voice AI to phone calls without running SIP equipment yourself, and it usually uses WebSocket and a telephone-network codec such as G.711. KAT Cloud provides a bidirectional voice stream, and you can also attach your own voice server via direct SIP connection.Related termsWebSocketSIP trunkSpeech-to-speech modelRead more on our blog (in Korean)
03
Voice AI
- Voice AI agent
- A voice AI agent is an AI that listens to and understands people's speech, answers by voice, and handles tasks such as taking reservations or looking up information. To work over the phone, it needs a phone number, lines and a real-time voice connection.Read more on our blog (in Korean)
- Speech-to-speech modelRealtime voice model
- A speech-to-speech model is a real-time voice model that takes speech as input and answers directly in speech without a separate text conversion step. KAT Cloud supports 14 voice models, including real-time voice models like these.Read more on our blog (in Korean)
- STT · Speech recognitionSpeech-to-Text
- STT (speech recognition) is technology that converts speech into text. KAT Cloud AI transcription keeps a text record of calls at KRW 10 per minute.Read more on our blog (in Korean)
- TTS · Speech synthesisText-to-Speech
- TTS (speech synthesis) is technology that reads text aloud in a human-like voice; it is the final stage in cascaded voice AI that chains STT, an LLM and TTS.Read more on our blog (in Korean)
- VADVoice Activity Detection
- VAD (voice activity detection) is technology that distinguishes speech from silence in audio; it is used to judge whether the other person has finished speaking and decide when the AI should answer.Read more on our blog (in Korean)
- Barge-in
- Barge-in is when the customer interrupts while the AI is speaking and the AI stops talking to listen. It is essential for natural phone conversations.Read more on our blog (in Korean)
- Turn-taking
- Turn-taking is how speakers trade turns in a conversation; a voice AI looks at silence length, intonation and context to judge whether the customer has finished speaking.Read more on our blog (in Korean)
- Response latencyLatency
- Response latency is the time from when the customer stops speaking until the AI's first audio is heard; it is the sum of network, voice activity detection, model inference and speech synthesis time.Read more on our blog (in Korean)
- HandoffAI-to-human handoff
- A handoff is passing a call that AI was handling to a human staff member. In KAT Cloud call records, calls connected to staff after AI handling are tagged AI_TO_DIRECT.Read more on our blog (in Korean)
- Echo cancellationAEC · Acoustic Echo Cancellation
- Echo cancellation (AEC) is signal-processing technology that removes the effect of sound from a speaker re-entering the microphone and reaching the other party as an echo.It matters especially for speakerphone and in-car calls. In voice AI, leftover echo can make the AI mistake its own voice for the customer's and stop talking, so it is closely tied to barge-in quality.Related termsBarge-inVADMOSRead more on our blog (in Korean)
- EndpointingEnd-of-speech detection
- Endpointing is the process of judging whether the customer has finished speaking, finalizing the speech recognition result and deciding when the AI should start answering.Looking only at silence length can cut the customer off even during a brief pause for breath, so methods that also consider intonation and sentence completeness are used. Judging too late increases response latency; judging too early cuts the customer off.Related termsVADTurn-takingResponse latencyRead more on our blog (in Korean)
- WER · Word error rateWord Error Rate
- WER (word error rate) is a metric that expresses speech recognition accuracy as the proportion of substituted, deleted and inserted words compared with the reference sentence.Because word spacing in Korean is inconsistent, the character error rate (CER) is often checked alongside it. Phone audio is 8 kHz narrowband, so even the same model tends to show a higher error rate than on ordinary recordings.Related termsSTT · Speech recognitionMOSG.711Read more on our blog (in Korean)
- LLMLarge Language Model
- An LLM (large language model) is an AI model trained on vast amounts of text to understand and generate language.In cascaded voice AI, the LLM understands the customer's words as transcribed by STT and writes an answer, which TTS then reads aloud. Real-time voice models handle speech directly but are built on the reasoning ability of a language model internally.Related termsSTT · Speech recognitionTTS · Speech synthesisSpeech-to-speech modelRead more on our blog (in Korean)
- System prompt
- A system prompt is an instruction given to an AI model before the conversation that sets its role, goals, tone and the rules it must follow.It determines what an AI agent should explain and what it must not do. In KAT Cloud, each agent has its own instructions (role, goals, tone) and reference information, and some rules always apply regardless of the instructions, such as never hiding that it is an AI.Related termsLLMVoice AI agentTool callingRead more on our blog (in Korean)
- Tool callingTool calling · Function calling
- Tool calling is how an AI model, during a conversation, picks a predefined external function (API), fills in its arguments, calls it, and uses the result in its answer.Typical examples are looking up delivery status by order number or checking available reservation times. A KAT Cloud AI agent can have up to 10 tools per agent, combining HTTP tools and MCP tools.Related termsMCPSystem promptVoice AI agentRead more on our blog (in Korean)
- MCPModel Context Protocol
- MCP (Model Context Protocol) is an open protocol that standardizes how AI models connect to external tools and data.Once an MCP server describes the tools it offers and their argument formats, the AI side can call tools on many servers in the same way. KAT Cloud AI agents can register an MCP server that uses Streamable HTTP as a tool and use that server's tools during calls.Related termsTool callingLLMVoice AI agentRead more on our blog (in Korean)
- IntakeStructured intake
- An intake is the record an AI agent produces by asking a customer for a defined set of fields, such as for a reservation or an order.In KAT Cloud, if you define the fields to collect in an agent's note-type (NOTE) manual, each time an intake is saved, changed or canceled it is delivered to your server immediately during the call as an INTAKE webhook. Unlike a call summary, it can be reflected in your booking ledger or order system before the call ends.Related termsWebhookVoice AI agentSystem promptRead more on our blog (in Korean)
04
Text messaging
- SMSShort Message Service
- SMS is a short text message of up to 90 bytes (about 45 Korean characters). KAT Cloud SMS starts at KRW 10 per message, with tiered rates that get lower as usage grows.Read more on our blog (in Korean)
- LMSLong Message Service
- LMS is a long text message, longer than SMS, that can have a subject line. The KAT Cloud chat API picks SMS, LMS or MMS automatically based on the length of the content.Read more on our blog (in Korean)
- MMSMultimedia Message Service
- MMS is a multimedia text message sent with an image. You upload the image first and then attach it to the send request.Read more on our blog (in Korean)
- KakaoTalk AlimTalk
- KakaoTalk AlimTalk is an informational message sent via KakaoTalk, using pre-approved templates for notices such as orders, reservations and deliveries. AlimTalk sending on KAT Cloud is coming soon and is not available yet.Read more on our blog (in Korean)
- KakaoTalk brand message
- A brand message is a message from a brand sent via KakaoTalk. Brand message sending on KAT Cloud is coming soon and is not available yet.Read more on our blog (in Korean)
05
API · Integration
- CPaaSCommunications Platform as a Service
- CPaaS is a cloud service that provides communications features such as calls and text messages as APIs so developers can add them directly to their own services. KAT Cloud is a CPaaS for AI agents.Read more on our blog (in Korean)
- Webhook
- A webhook is a way for a service to notify you of an event by sending an HTTP POST request to a server address you registered in advance. KAT Cloud sends CALL_RECEIVED, CALL_COMPLETED, MESSAGE_RECEIVED and SEND_RESULT events.Read more on our blog (in Korean)
- API keyx-api-key
- An API key is a secret string that identifies the user sending an API request. The KAT Cloud API authenticates requests with a key issued in the console, sent in the x-api-key header.Read more on our blog (in Korean)
- Group
- A group is the operating unit of KAT Cloud, used to separate stores, branches or projects. You get a group with one phone number (channel) when you sign up, and you can attach more numbers by API or MCP. Billing and usage are also managed per group.Read more on our blog (in Korean)
- Channel
- A channel is a phone number connected to a group. Calls and text messages are exchanged through this number, and a group can have multiple channels.Read more on our blog (in Korean)
- Chat room
- A chat room is a conversation thread created from the combination of a channel (your number) and a customer's phone number. The same combination always uses the same chat room, so the message history continues.Read more on our blog (in Korean)
- Rate limit
- A rate limit is the maximum number of API requests that can be sent in a given period. The KAT Cloud API allows 1,200 requests per 60 seconds.Read more on our blog (in Korean)
- Incremental sync
- Incremental sync is fetching only the data that changed since the last fetch. KAT Cloud call records use the updatedAfter value to fetch only records updated after the last time you received them.Read more on our blog (in Korean)
- Presigned URL
- A presigned URL is a temporary, signed address that allows a file to be downloaded only for a limited time. KAT Cloud call recording file URLs expire after 1 hour.Read more on our blog (in Korean)
- HMAC signatureHMAC-SHA256
- An HMAC signature is a method of computing a hash from a secret and a message to verify that the sender holds the secret and that the content has not been altered.KAT Cloud webhooks concatenate X-Webhook-Timestamp with the raw body, compute HMAC-SHA256 with your account's secret, and send the result in the X-Webhook-Signature header in v1= format. The receiving server performs the same calculation, compares the values in constant time, and rejects requests whose timestamp differs by more than 5 minutes.Related termsWebhookIdempotencyAPI keyRead more on our blog (in Korean)
- Idempotency
- Idempotency is the property that processing the same request or notification several times has the same result as processing it once.Webhooks are resent when the response is slow or fails, so the same notification can arrive twice. KAT Cloud webhooks keep the same eventId on retries, so the receiving server can prevent duplicate processing by recording the eventIds it has handled.Related termsWebhookHMAC signatureIncremental syncRead more on our blog (in Korean)
06
Regulation · Privacy
- Caller ID spoofing
- Caller ID spoofing is disguising the number shown to the recipient of a call or text message as a number other than the real one.Because it is abused in voice phishing that impersonates public institutions or financial companies, it is prohibited under Korean law, and carriers block or correct calls with falsely displayed numbers. Calls made by AI must also display only numbers that have been assigned to the caller or that the caller is authorized to use.Related termsCaller IDCaller ID pre-registrationInternational call noticeRead more on our blog (in Korean)
- International call notice
- An international call notice is a measure by carriers that tells the recipient that a call is coming from overseas.To reduce scam calls that are made from overseas while posing as Korean numbers, carriers attach a notice to calls originating abroad or block spoofed numbers. If you call Korean customers from overseas servers over overseas lines, this notice is attached and can reduce trust, so services for Korea usually call over Korean lines and numbers.Related termsCaller ID spoofingCaller IDPublic switched telephone networkRead more on our blog (in Korean)
- Opt-out
- Opt-out refers to a recipient's expressed wish not to receive advertising calls or text messages, and the procedure for handling it.In Korea, when you send advertising information you must also tell recipients how they can refuse or withdraw consent free of charge, and you must not send more to anyone who has refused. Sending systems should add refusal requests to the exclusion list immediately.Related termsCaller ID spoofingSMSAI disclosureRead more on our blog (in Korean)
- Communication confirmation dataCommunication metadata
- Communication confirmation data is not the content of calls but records of the fact that communication took place, such as call date and time, the other party's number, units used and access logs.Korea's Protection of Communications Secrets Act sets out the procedure by which investigative agencies and others can request this data, and the obligation of telecommunications businesses to retain it for a set period. It is legally distinct from communication content such as call recordings, so retention and disclosure rules must be considered separately.Related termsCall unitsOutsourcing of personal data processingFacilities-based telecommunications carrierRead more on our blog (in Korean)
- Outsourcing of personal data processingProcessing outsourcing
- Outsourcing of personal data processing is when a personal information controller entrusts another company with processing personal information for its own business.In AI phone services, call data is often entrusted to providers of communications infrastructure, speech recognition, language models and CRM. Korea's Personal Information Protection Act requires the outsourcing terms to be set out in writing and the processors and outsourced tasks to be disclosed, and processors may handle the information only within the scope entrusted to them.Related termsCross-border transfer of personal dataCommunication confirmation dataSTT · Speech recognitionRead more on our blog (in Korean)
- Cross-border transfer of personal dataCross-border data transfer
- A cross-border transfer of personal data is providing personal information to a company located overseas, or entrusting it to be processed or stored overseas.Sending call audio and transcripts to speech recognition or language models in overseas regions may count as a cross-border transfer. Korea's Personal Information Protection Act allows cross-border transfers only when set requirements are met, such as separate consent or disclosure in the privacy policy, so you should document the items, countries and recipients involved.Related termsOutsourcing of personal data processingLLMSTT · Speech recognitionRead more on our blog (in Korean)
- AI disclosure
- AI disclosure is informing users that they are being served by an AI rather than a person, or that a result was generated by AI.Korea's AI Basic Act, in force since January 2026, requires that users be notified in advance when a service is based on generative AI. On a phone call with no screen, the most reliable approach is for the AI agent to say it is an AI in its first greeting. KAT Cloud AI agents never hide that they are AI regardless of their instructions, and answer truthfully when asked whether they are human.Related termsVoice AI agentOpt-outTTS · Speech synthesisRead more on our blog (in Korean)
Get started
Give the AI you built today
a phone number
Use the Personal plan free for 7 days after you sign up. Get a number and take your first call.
Already have an account? Log in
The console (app.katcloud.kr) is in Korean.
- 01Create an account
- 02Get a phone number
- 03Connect your agent and take the first call
