Start here
Getting started
Create an account, pick your models, connect your tools, and have the assistant build your first workflow.
Murmurator turns plain-language descriptions into working automations. You describe what should happen; the workflow assistant designs the workflow, checks it against your real connections and models, saves it, and helps you test it. Every run is then visible live, step by step.
1. Create your account
Sign up with your name, work email and team name. You become the account's owner and your free trial starts immediately — no card required.
Invite teammates from Members. Roles decide what people can change:
| Role | Can do |
|---|---|
| Owner | Everything, including billing, spend limits and promoting other owners |
| Admin | Manage connections, AI models and members |
| Member | Build, edit and run workflows; view everything |
A person can belong to several accounts and switch between them from the sidebar.
2. Choose your models
Workflows that think — summarizing, reviewing, classifying, answering — use AI models. You have two options, and you can mix them:
- Murmurator AI — built-in Claude models (Fable, Opus, Sonnet, Haiku). Click Use under AI models → Murmurator AI and you're done. Usage is billed per token; trials include free credit.
- Your own keys — add a key for Anthropic, OpenAI, Gemini, OpenRouter, Mistral, DeepSeek, xAI or Ollama. You pay your provider directly.
Either way, you give each model a short name like smart or fast. Steps refer to that name, and one model is the account default. See AI models.
3. Connect your systems
Under Connections, add GitHub, Slack, Linear, Stripe, Zendesk, a database, an email provider or any of the others on the integrations page. Each connection gets a slug (for example github or warehouse) that workflows use to refer to it, and exposes a set of tools such as github.create_comment or warehouse.query. Credentials are encrypted and never appear in workflows. See Connections.
4. Describe your workflow
From the overview page or Workflows, describe the automation:
When a pull request is opened in acme/api, review the diff and post a summary comment. Alert #eng-reviews if it looks risky.
A draft workflow is created and the assistant starts working in the panel beside it. It may ask a clarifying question — which channel, which repository — then saves the first version. The steps diagram appears as soon as it does. See Workflows for everything the assistant can do.
5. Try it, then turn it on
Press Run to start a manual run and watch each step execute: its input, output, logs, model output streaming in, and token usage. When you're happy, press Enable trigger (or ask the assistant to) so the workflow runs automatically on its trigger.
Keep improving it by chatting: "also post to #releases", "skip draft PRs", "why did the last run fail?". Every change becomes a new version you can review.