# Voice.GradientFlow

## Docs

- [What is Voice.GradientFlow?](/getting-started/introduction): Build a voice agent that answers calls for your business. Describe it, test it in your browser, publish it with one button.
- [Build your first agent](/getting-started/first-agent): Start from a template or a short description, test the agent in the Playground, then publish it and share a link.
- [Keys](/getting-started/keys): What your workspace already provides, when you need a key of your own, and where to enter it.
- [Your own model server](/getting-started/own-model-server): Run the agent's model on Ollama, vLLM, LiteLLM, LM Studio, a gateway, or any server that speaks the OpenAI API, with or without a key.
- [Around the builder](/building-flows/around-the-builder): Where every panel, rail and button lives: the Simple and Advanced views, toolbar, left rail, right rail, canvas, node editor, agents list and version history.
- [Node reference](/building-flows/node-overview): Every node in the Components panel, what it does, every setting with its help text, what it connects to and which key it needs.
- [Skills and routing](/building-flows/conversation-nodes): How the Persona, Skills, branches and End fit together, and how the agent decides which Skill to use next.
- [Build with AI](/ai-generation/ai-builder): Describe the agent in plain language. The AI builder asks a few questions, builds it on the canvas, and makes changes when you ask.
- [Playground](/testing/playground): Talk to your agent in the browser. Chat modes, input variables, suggestions, transcript, diagnostics and history.
- [Watch the call on the canvas](/testing/step-highlighting): During a Playground call the canvas shows which Skill is active, which branch was taken and which tool is running.
- [Troubleshooting](/testing/troubleshooting): What each readiness message and Playground error means, and what to do about it.
- [MCP servers](/integrations/mcp-servers): Connect a Model Context Protocol server so the agent can use its tools during a call.
- [Web search (Exa)](/integrations/web-search): Let the agent look things up on the web during the call.
- [Calendar booking (Cal.com)](/integrations/calendar-booking): Let the agent check free slots and book appointments in your Cal.com calendar.
- [CRM (HubSpot)](/integrations/crm): Look up and update HubSpot contacts during the call, and sync every call to HubSpot afterwards.
- [SIP transfer](/integrations/sip-transfer): Let the agent transfer a live call to a real phone number.
- [Structured output](/integrations/structured-output): Pull named fields out of the finished conversation, such as name, email and whether the issue was resolved.
- [Webhooks](/integrations/webhooks): Send the call data to your own system when the call ends, and receive signed call and campaign events.
- [Helpdesk tickets](/integrations/helpdesk): Open a follow-up ticket after every call, in HubSpot Tickets or in this workspace's Dashboard.
- [Publish and share](/deployment/publish-and-share): Publish a version of your agent, get a share link, make links with your own expiry and call limit, revoke them, embed one on a website, and understand what happens when you publish again.
- [Export and import](/deployment/export-json): Save an agent as a file to back it up or move it, and import the file onto an empty canvas.
- [Installation](/deployment/installation): Operator guide: run your own Voice.GradientFlow on one machine with Docker for the media server and worker, and Node.js for the app.
- [Docker Compose](/deployment/docker): What each Compose file in the repository starts, which commands the containers run, and how the hosted environments are laid out.
- [Python Worker](/deployment/python-worker): Run the agent worker natively with uv instead of in Docker, for faster iteration on the Python side.
- [Agent Runtime](/deployment/agent-runtimes): How the Voice.GradientFlow Python agent worker runs alongside the Next.js app.
- [Environment variables](/deployment/environment-variables): Operator reference for the server environment: voice infrastructure, AI builder provider, orchestration default, provider keys, runtime limits, health endpoints and metrics.