Skip to main content
Connect your voice provider, run one test conversation, and watch it land in Zelto as a transcribed, analyzed conversation. Start to finish, about five minutes — and none of it changes how your agents run.
1

Create an account

Sign up at dashboard.zelto.ai/signup with your email or Google. Email sign-up sets a password and emails a 6-digit code — paste it back to verify.
2

Create or join an organization

After verifying, name your organization to create it — your URL becomes dashboard.zelto.ai/org/<slug>. If a teammate already invited you, accept their invitation to join that workspace.An organization is your workspace and account boundary. If your team operates agents for several clients or brands, model those inside the workspace as companies, not as separate organizations.
3

Connect your voice provider

Open Get started, choose the stack your agents run on, and paste its API key:
  • Vapi, Retell, and ElevenLabs — connect the account, import an agent on the next step, and Zelto starts pulling its conversations automatically. You can add an optional webhook later under Settings → Integrations for lower-latency delivery.
  • LiveKit or any other stack — Bland, Pipecat, in-house — mint an API key and POST each finished conversation to /webhooks/calls. See Custom & other providers.
Not sure which is yours? Connect a voice provider has a one-table decision guide.
4

See your first conversation

Place a test call to your voice agent (or message your text agent) — or POST one to /webhooks/calls if you took the custom path. The next provider sync adds a new conversation with the transcript and, when present, the recording; an optional webhook makes it arrive within seconds.
Nothing showing up? Open the agent and check its Connections card — it shows when the last conversation arrived, or flags an error, so you can confirm at a glance that conversations are reaching Zelto. See Connect a voice provider for the full check.
5

See what Zelto finds

Zelto analyzes each conversation as it arrives. What you can do next depends on your plan. Every organization starts on the Free plan, which includes conversations, transcripts and traces.
  • On Observability, create monitors that judge every conversation against a rule you write, and trend the result over time.
  • On Improve, Zelto also groups the same problem across conversations into findings — a recurring script failure, a mispronounced product name, a tool the agent keeps misusing — so you don’t have to read every transcript yourself.
With findings, you can go further:
  • Invite teammates from Settings → Members so the right people see each finding — see Manage your team.
  • If you serve several clients, set up companies to attach their agents and attribute every new conversation to the right client.
  • Open a finding to see the conversations behind it and read the exact transcript moments that triggered it.
  • Push any finding to Linear from the finding’s panel to hand it to the team that owns the fix.

Where to go next

  • Send traces — export spans, identify agents, and link conversations.
  • Onboard with an agent — skip the clicks: have your coding agent connect a provider, create a monitor, and route findings over MCP.
  • How Zelto works — the whole pipeline, from conversation to finding to fix.
  • Connect Slack — chat with Zelto and get each agent’s daily status posted to a channel.
  • Script against your data over the REST API or MCP, using a key minted under Settings → Integrations → API integration.