# Kilevel > Kilevel is an AI agent platform for customer conversations and the work behind them: voice agents that answer and place phone calls, chat and messaging agents, a visual automation canvas, and the CRM, analytics and telephony around them. This file is a map of kilevel.com written for language models, in the llms.txt format (https://llmstxt.org). Every link is absolute so it can be fetched on its own. The full text of the documentation, concatenated into one file, is at https://kilevel.com/llms-full.txt. Every URL the site publishes is at https://kilevel.com/sitemap.xml. The dashboard and other signed-in surfaces are not listed here and are disallowed in robots.txt. How an agent obtains a credential for the API and the MCP server, and how to present it, is at https://kilevel.com/auth.md. Every page listed below is served as Markdown to anything that asks for it: send `Accept: text/markdown`, or append `.md` to the path. HTML remains the default, so a browser is unaffected. ## Product - [Kilevel](https://kilevel.com): Product overview: AI voice agents for inbound and outbound calls, a chat widget, WhatsApp, SMS and email, plus the automation canvas and CRM behind them. - [Pricing](https://kilevel.com/pricing): Plans, what each tier includes, and how call, chat and automation usage is metered. - [Research](https://kilevel.com/research): The research agenda for agent-to-agent work: capability marketplaces, an agent identity layer, MCP coordination, self-improving agents. Each thread is marked with its stage, so read it as direction rather than shipped product. - [Documentation](https://kilevel.com/docs): Guides, integration notes, the developer reference and the API reference. Every page is listed individually further down this file. ## Use cases - [Any use case](https://kilevel.com/any-use-case): The general path from idea to a live agent, for teams whose case is not one of the ones below. Every job is built from the same blocks — answer, ask, look up, hand over, follow up — and one build works on the phone, the website widget, WhatsApp, SMS and email. - [Agents and the State Graph](https://kilevel.com/use-cases/agents): How a Kilevel agent is built: the State Graph node catalogue, stations, evaluations and tool-enforced validation — prompt-driven behaviour inside deterministic gates. - [Appointment booking](https://kilevel.com/use-cases/appointment-booking): Availability rules wired into voice and chat flows: slot search, booking intents, SMS and email confirmation. - [Automations](https://kilevel.com/use-cases/automations): The workflow canvas: AI inference, data stores, HTTP and MCP calls, sandboxed code, media generation, call campaigns and human approval steps. - [Customer support](https://kilevel.com/use-cases/customer-support): Inbound support across the web widget, chat and phone, with backend lookups during the conversation. - [Lead generation](https://kilevel.com/use-cases/lead-generation): Outbound call campaigns: list handling, lead prioritisation, retries, and what happens to a qualified lead. - [Performance analytics](https://kilevel.com/use-cases/performance-analytics): Transcription and conversation analysis for calls and chats run on Kilevel, and for recordings pushed in over API, FTP/SFTP, Google Drive, S3-compatible storage or upload. Metric packs, custom metrics, weighted scorecards, KPI boards and alerts. - [Softphone and contact centre](https://kilevel.com/use-cases/softphone-contact-center): The browser softphone, whisper channels, live handoff from an agent to a human operator, and per-session analytics. - [Surveys and feedback](https://kilevel.com/use-cases/surveys-feedback): Voice interviews for CSAT and feedback collection, with structured answers parsed out of the transcript. - [Work agents](https://kilevel.com/use-cases/work-agents): Work mode: flows running autonomously in the background with their own sandbox VM, recorded browser sessions, sub-agent orchestration, deliverables, schedules, budgets and approvals. ## Industries - [Any industry](https://kilevel.com/any-industry): The general path from idea to a live agent, whatever the line of business. The model has already read the trade; the setup is what it must never say, what to write down, when to fetch a person, and what the local rules make it say first — plus the consent, disclosure and retention switches behind that. - [E-commerce](https://kilevel.com/industries/ecommerce): Order tracking, product questions and post-purchase support. - [Education](https://kilevel.com/industries/education): Course enquiries, enrolment and student scheduling for schools and universities. - [Energy and utilities](https://kilevel.com/industries/energy): Billing enquiries, meter readings and service calls. - [Finance](https://kilevel.com/industries/finance): Caller verification, account enquiries and scheduling under GDPR constraints. - [Healthcare](https://kilevel.com/industries/healthcare): Patient booking, reception overflow and reminders for clinics and practices. - [Insurance](https://kilevel.com/industries/insurance): First-notice-of-loss intake, policy questions and lead qualification. - [Real estate](https://kilevel.com/industries/real-estate): Property lead capture, viewing scheduling and tenant enquiries. - [Travel and tourism](https://kilevel.com/industries/travel-tourism): Booking changes, reservation status and multilingual traveller support. ## Company - [Trust Center](https://kilevel.com/trust-center): GDPR posture in detail: controller and processor roles, the DPA, the full subprocessor list, security measures, and which controls the customer operates rather than Kilevel. - [Academy](https://kilevel.com/academy): The certification ladder for people who design and operate Kilevel deployments, from Platform Operator to Forward-Deployed AI Engineer. The lessons themselves are behind a login and are not in this file. - [About](https://kilevel.com/company/about): What Kilevel is, who builds it, and where the company is registered. - [Partner programme](https://kilevel.com/company/partner): Terms for agencies and integrators reselling or delivering on Kilevel. - [Careers](https://kilevel.com/company/careers): Open roles and how the team works. ## API reference - [API reference](https://kilevel.com/docs/api-reference): REST API, grouped by resource. Base URL https://api.kilevel.com. Authentication is an X-API-Key header. - [Account & Utility](https://kilevel.com/docs/api-reference/account-utility): 6 endpoints. - [Agent Knowledge Base](https://kilevel.com/docs/api-reference/agent-knowledge-base): 4 endpoints. - [AI Flows & Agents](https://kilevel.com/docs/api-reference/ai-flows-agents): 6 endpoints. - [Automation Artifacts](https://kilevel.com/docs/api-reference/automation-artifacts): 2 endpoints. - [Call & Chat History](https://kilevel.com/docs/api-reference/call-chat-history): 8 endpoints. - [Call & Chat Injections](https://kilevel.com/docs/api-reference/call-chat-injections): 2 endpoints. - [Call Jobs](https://kilevel.com/docs/api-reference/call-jobs): 7 endpoints. - [Custom Tools](https://kilevel.com/docs/api-reference/custom-tools): 5 endpoints. - [Data Stores](https://kilevel.com/docs/api-reference/data-stores): 11 endpoints. - [Outbound Triggers](https://kilevel.com/docs/api-reference/outbound-triggers): 2 endpoints. - [Structured Outputs](https://kilevel.com/docs/api-reference/structured-outputs): 4 endpoints. - [Test Endpoints](https://kilevel.com/docs/api-reference/test-endpoints): 10 endpoints. - [Transcriptions](https://kilevel.com/docs/api-reference/transcriptions): Push recordings into a transcription job and read back their transcripts, analysis and attached details. (6 endpoints) - [Usage Reports](https://kilevel.com/docs/api-reference/usage-reports): 1 endpoint. ## Docs: Getting started - [What is Kilevel](https://kilevel.com/docs/getting-started/introduction): A voice AI platform for building agents that answer and place real phone calls, run the conversation against your own data, and hand off to the rest of your stack. - [Quickstart](https://kilevel.com/docs/getting-started/quickstart): Place your first outbound call from the API in four steps — key, flow, call, result. - [Core concepts](https://kilevel.com/docs/getting-started/core-concepts): Flows, agents, call jobs, applications, data stores and phone numbers — what each object is and which one you actually want. ## Docs: API guides - [Authentication](https://kilevel.com/docs/api/authentication): Public API requests carry either an X-API-Key header or an OAuth2 bearer token. What each is scoped to, and what a rejected request looks like. - [Rate limits](https://kilevel.com/docs/api/rate-limits): 120 requests per minute per key by default, reported on every response, with a 429 and a reset when the bucket empties. - [Outbound calls](https://kilevel.com/docs/api/outbound-calls): Queue one call with POST /outbound/calls — by saved flow, by workspace default, or with a one-off flow defined inline. - [OAuth2 client credentials](https://kilevel.com/docs/api/oauth-clients): Exchange a client_id and client_secret for a short-lived bearer token that calls the public API and the MCP server with exactly the permissions you granted. ## Docs: Build - [Flows and agents](https://kilevel.com/docs/build/flows-and-agents): The flow holds what a conversation does; the agent decides how it sounds. How the two fit together, and how a flow is structured. - [The State Graph](https://kilevel.com/docs/build/flows-and-agents/state-graph): Lay a flow out as a graph — where the conversation is, what it decides, what happens along the way, and how it ends. - [States](https://kilevel.com/docs/build/flows-and-agents/state-graph/states): Named conditions a session is in — each with a typed value and an on/off flag — that steer the prompt and decide what the assistant may do. - [Nested states](https://kilevel.com/docs/build/flows-and-agents/state-graph/states/nested-states): Organise states as a tree up to three levels deep — modes with sub-modes — with values that cascade from parent to child. - [Collecting information](https://kilevel.com/docs/build/flows-and-agents/state-graph/collecting-information): Nodes that hold the conversation on one question until the answer is valid — typed values, keypad entries and recording consent. - [Decisions](https://kilevel.com/docs/build/flows-and-agents/state-graph/decisions): Grade the conversation with evaluations, answer several at once with a group, and apply named transitions with Change State. - [Actions](https://kilevel.com/docs/build/flows-and-agents/state-graph/actions): Nodes that happen in passing — speak a line, run a tool, steer the assistant quietly, write memory, notify someone, start an automation. - [Handoffs and ending](https://kilevel.com/docs/build/flows-and-agents/state-graph/handoffs-and-ending): Hand the session to a person, another agent or another flow — or end it cleanly, with a closing line that always completes. - [Events, jumps and notes](https://kilevel.com/docs/build/flows-and-agents/state-graph/events-jumps-and-notes): React to silence and time passing, keep a large graph readable, and document it for the next person who opens it. - [Voice and language](https://kilevel.com/docs/build/voice-and-language): Speech recognition, voices, languages and pronunciation — and how the assistant handles interruptions and silence. - [Knowledge base](https://kilevel.com/docs/build/knowledge-base): Give an agent documents, web pages and text to answer from — shared across agents or kept local to one. - [Tools](https://kilevel.com/docs/build/tools): Let the assistant act during a conversation — built-in tools, your own HTTP endpoints, call transfers and long-running asynchronous work. - [Custom HTTP tools](https://kilevel.com/docs/build/tools/custom-http-tools): Describe your own endpoint as a tool the assistant can call — parameters it fills from the conversation, and a response it reads back. - [Asynchronous tools](https://kilevel.com/docs/build/tools/asynchronous-tools): Tools that report back later — how a result reaches a conversation still in progress, and what the assistant does with it. - [Call transfers](https://kilevel.com/docs/build/call-transfers): Hand a call to a phone number, a SIP destination or another flow — blind, or warm with a briefing for the person picking up. - [Structured outputs](https://kilevel.com/docs/build/structured-outputs): Turn every conversation into fields your systems can use — defined once as a schema, filled after each call or chat. ## Docs: Launch - [Phone numbers](https://kilevel.com/docs/launch/phone-numbers): Connect numbers through Twilio, Telnyx or your own SIP trunk, decide what each is for, and set how many calls it may carry at once. - [Inbound calls](https://kilevel.com/docs/launch/inbound-calls): Decide what happens when a number rings — the agent answers, the call rings through, is rejected, forwarded, or handed to your own system. - [Campaigns](https://kilevel.com/docs/launch/campaigns): Work through a list of contacts with call jobs — call windows, pacing, retries and results, from the dashboard or the API. - [Website widget](https://kilevel.com/docs/launch/website-widget): Put your agent on your website — chat, or a voice call straight from the browser — with one script tag. - [Messaging channels](https://kilevel.com/docs/launch/messaging): Run the same flows over WhatsApp, SMS, email and Slack — and start conversations yourself from the API. ## Docs: Monitor & improve - [Call and chat history](https://kilevel.com/docs/monitor/call-history): Every conversation leaves a record — transcript, recording, outcome and extracted fields — to review, export or pull over the API. - [Live monitoring](https://kilevel.com/docs/monitor/live-monitoring): Listen to calls as they happen, follow live captions, and step in — by injecting speech, or by taking the call yourself. - [Transcription](https://kilevel.com/docs/monitor/transcription): Transcribe audio recorded outside Kilevel — uploads, cloud storage, FTP, Google Drive or URLs — and run the same analysis a call gets. - [Sources and watching](https://kilevel.com/docs/monitor/transcription/sources): Where a transcription job reads audio from — buckets, file servers, Drive, uploads, URLs or your own API — and how a job keeps running once the first batch is done. - [Analysis and metric packs](https://kilevel.com/docs/monitor/transcription/analysis): What the model extracts from each finished transcript — ready-made metric packs, your own structured outputs, written reasons, and the metrics measured straight from the audio. - [Details and linking](https://kilevel.com/docs/monitor/transcription/details): Attach the agent, customer, project and conversation time your own systems already hold — and tie imported recordings to the agents and contacts in your workspace. - [Pushing recordings over the API](https://kilevel.com/docs/monitor/transcription/api): Hand recordings to a standing transcription job from your own system, and read the transcript, analysis and details back. - [Performance](https://kilevel.com/docs/monitor/performance): How every agent is doing, across calls, chats and imported recordings — scorecards, trends, alerts, calibration, and the privacy settings a works council agreement needs. - [Evaluations and benchmarks](https://kilevel.com/docs/monitor/evaluations): Test a flow against scenarios before it meets real callers, and compare models and settings with benchmarks. - [Automations](https://kilevel.com/docs/monitor/automations): Workflows that run around your conversations — triggered by calls, schedules or webhooks, and connected to the rest of your stack. - [Triggers](https://kilevel.com/docs/monitor/automations/triggers): The five ways an automation starts — by hand, on a schedule, from a webhook, as a tool, or from an event on the platform. - [AI conversations](https://kilevel.com/docs/monitor/automations/ai-conversations): Start an outbound AI call or chat from an automation, then wait for the outcome — or watch the conversation and react while it runs. - [Collecting input from people](https://kilevel.com/docs/monitor/automations/collecting-input): Pause an automation and ask a person for text, files or media — over a link, the inbox, email, a Work channel or the API. ## Docs: Webhooks - [Session webhooks](https://kilevel.com/docs/webhooks/session-webhooks): Get a request when a call or chat starts, when it ends, and when its analysis is ready — for every flow, or per flow. - [Call job completion webhook](https://kilevel.com/docs/webhooks/call-job-completion): One request when every row of a call job has finished — with the counts, and the URLs to fetch the rows. ## Docs: MCP (Developers) - [Kilevel as an MCP server](https://kilevel.com/docs/mcp-server/kilevel-mcp-server): Use Kilevel from Claude, ChatGPT, IDEs and any other MCP client — list flows, read call history and trigger calls with the same API key. ## Docs: Integrations - [Integrations overview](https://kilevel.com/docs/integrations/overview): Connect calendars, workspace tools, messaging, AI and voice providers, telephony and MCP servers — once, for every flow and automation. ## Docs: Scheduling - [Calendars and booking](https://kilevel.com/docs/scheduling/calendars): Let an agent check availability and book appointments in Google Calendar, Cal.com, Calendly or Kilevel's own booking calendar. ## Docs: Productivity - [Google Workspace](https://kilevel.com/docs/productivity/google-workspace): Read and write Google Sheets, send email through Gmail, and use Google Drive files — from flows, automations and transcription jobs. - [Notion](https://kilevel.com/docs/productivity/notion): Use a Notion workspace from automations — look up pages and databases, and write results back. - [Slack](https://kilevel.com/docs/productivity/slack): Post call results and alerts to Slack channels, and hold conversations with your flows inside Slack. - [Email and SMS](https://kilevel.com/docs/productivity/email-and-sms): Send email through your own mail server and text messages from an SMS-enabled Twilio number. ## Docs: AI & voice providers - [AI model providers](https://kilevel.com/docs/models/ai-providers): Bring your own keys for OpenAI, Anthropic, Gemini, Azure OpenAI and a dozen more — for conversations, automations and media generation. - [Voice providers](https://kilevel.com/docs/models/voice-providers): Use your own ElevenLabs, PlayHT or Deepgram accounts for voices and speech recognition. ## Docs: Telephony - [SIP trunks](https://kilevel.com/docs/telephony/sip-trunks): Bring your own carrier or PBX — connect a SIP trunk for inbound and outbound calls, with separate outbound credentials if your provider needs them. - [Twilio and Telnyx](https://kilevel.com/docs/telephony/twilio-and-telnyx): Use numbers from your own Twilio or Telnyx account for voice, SMS and WhatsApp. - [WhatsApp](https://kilevel.com/docs/telephony/whatsapp): Connect WhatsApp through Twilio or the Meta WhatsApp Cloud API, and run your flows on it. ## Docs: MCP (Integrations) - [Connecting MCP servers](https://kilevel.com/docs/mcp/mcp-servers): Give flows and automations the tools and resources of any Model Context Protocol server. ## Docs: Workspace - [Team and roles](https://kilevel.com/docs/workspace/team-and-roles): Invite your team and give each person the access they need — from full workspace control to a call-centre portal. - [API keys and permissions](https://kilevel.com/docs/workspace/api-keys): Create keys with full access, or limit a key to exactly the endpoints an integration needs. ## Docs: Billing & usage - [Plans and billing](https://kilevel.com/docs/billing/plans-and-billing): Choose a plan, manage your subscription and payment method, and download invoices. - [Balance and top-ups](https://kilevel.com/docs/billing/balance-and-top-ups): How usage is paid from your balance — topping up, automatic top-ups, and your transaction history. - [Concurrency and limits](https://kilevel.com/docs/billing/concurrency-and-limits): How many calls, chats and automations can run at once — set by your concurrency lines, and visible on the limits overview. ## Docs: Agencies - [Agencies and white-label](https://kilevel.com/docs/agencies/agencies-and-white-label): Run Kilevel for your clients — as a partner, distributor or reseller — with hosted pages on your own domain. ## Docs: Security & compliance - [Data protection](https://kilevel.com/docs/compliance/data-protection): Where to find the GDPR documentation, the subprocessor list and data locations — and the controls you operate yourself. - [AI disclosure](https://kilevel.com/docs/compliance/ai-disclosure): Tell callers they are speaking with an AI at the start of every call — preset wording in over 30 languages, or your own. - [Recording consent](https://kilevel.com/docs/compliance/recording-consent): Start recording only after the caller has agreed — with a consent step that asks, checks the answer, and starts the recording itself. ## Changelog - [Changelog](https://kilevel.com/docs/changelog): Customer-facing release notes, newest first. The entries below are anchors into this one page. - [Smart-MCP](https://kilevel.com/docs/changelog#2026-10-05-smart-mcp): 2026-10-05 - [Training modules with personas](https://kilevel.com/docs/changelog#2026-10-03-training-modules-personas): 2026-10-03 - [Script Adherence](https://kilevel.com/docs/changelog#2026-10-03-script-adherence): 2026-10-03 - [Guides for transcription and Performance](https://kilevel.com/docs/changelog#2026-09-22-transcription-and-performance-docs): 2026-09-22 - [Performance](https://kilevel.com/docs/changelog#2026-09-21-performance): 2026-09-21 - [Transcribe straight from your own storage](https://kilevel.com/docs/changelog#2026-09-20-transcription-sources): 2026-09-20 - [Metric packs](https://kilevel.com/docs/changelog#2026-09-20-metric-packs): 2026-09-20 - [Details on every recording, and an API to push them](https://kilevel.com/docs/changelog#2026-09-19-recording-details-and-api): 2026-09-19 - [Steadier Works, and artifacts you can actually read](https://kilevel.com/docs/changelog#2026-09-18-works-sandboxes-artifacts): 2026-09-18 - [Test runs stay private](https://kilevel.com/docs/changelog#2026-09-15-private-test-runs): 2026-09-15 - [Jump nodes and connector styles](https://kilevel.com/docs/changelog#2026-09-15-jumps-and-connectors): 2026-09-15 - [A new documentation site](https://kilevel.com/docs/changelog#2026-09-14-documentation): 2026-09-14 ## Case studies - [Case studies](https://kilevel.com/case-studies): Deployments in production: what was automated, how it was built, and what changed. - [Project N, a German outbound call center](https://kilevel.com/case-studies/project-n-outbound-transformation): An established call center with 70 to 100 agents moved its high-volume outbound campaigns onto autonomous voice agents, then rebuilt the operation around them. - [Car Dealership S](https://kilevel.com/case-studies/car-dealership-s-call-overflow): A 15-person team fielding more than 700 inquiries a day was losing several hundred calls daily to overflow. Agents now take everything the team cannot. - [A B2B software company in Bavaria](https://kilevel.com/case-studies/b2b-software-bavaria-demo-booking): Outbound demo scheduling for trial users, first-level technical support on inbound, and a clean handover to human experts when the question gets hard. - [An e-commerce cosmetics brand](https://kilevel.com/case-studies/ecommerce-cosmetics-cart-recovery): Round-the-clock product consultation, order tracking answered in the call, and outbound follow-up that goes after abandoned carts. ## Blog - [Blog](https://kilevel.com/blog): Engineering and compliance notes from the team building the platform. - [Article 50 puts a sentence at the front of every call](https://kilevel.com/blog/eu-ai-act-article-50-greeting): The EU AI Act obliges you to tell callers they are speaking to a machine. That obligation lands in one specific place, the greeting, and it changes how the first three seconds of a call have to be built. - [A longer prompt is not a conversation design](https://kilevel.com/blog/state-graph-not-a-longer-prompt): Every voice agent that fails in production fails the same way: the prompt grew until the model could no longer hold all of it at once. A state graph is what you reach for when adding instructions stops working. - [Where the second goes in an AI phone call](https://kilevel.com/blog/where-the-second-goes-latency): Response latency is the single loudest quality signal in a voice agent, and it is almost never one slow component. It is five costs in sequence, and most of the wins come from overlapping them rather than speeding any one of them up. ## Optional - [Kilevel (Deutsch)](https://kilevel.com/de): German-language homepage. Most other pages serve German from the same URL. - [Imprint](https://kilevel.com/legal/imprint): Company registration, address and responsible contact. - [Privacy policy](https://kilevel.com/legal/privacy): What personal data the platform processes, on what basis, and for how long. - [Terms of service](https://kilevel.com/legal/terms): The contract governing use of the platform. - [Refund policy](https://kilevel.com/legal/refund): When wallet balance and subscription fees are refundable. - [Data Processing Agreement](https://kilevel.com/legal/dpa): The Article 28 GDPR processor agreement, including subprocessors and transfer mechanisms. - [Acceptable Use Policy](https://kilevel.com/legal/aup): What the platform may not be used for, including calling conduct and consent requirements.