Skip to content

Flows

AI agents and graphs: from the home tour and first agent to node setup, testing, triggers, API, publishing, quality evaluation, and operator handoff.

Start here

Getting to know Flows: home and agent catalog

Getting started

Chat with Lara — the platform assistant Create your first agent from a template

Graph editor

Studio and canvas: two editor modes Branch the dialog with edge conditions Shape agent answers: prompt and model Keep a large graph readable

Agent capabilities

Give the agent tools Add an answer self-review Let the agent call an operator itself Hand off the dialog between agents Agent files a background task for the team Agent step plan via Skill Agent step plan via tools Attach a Skill to the agent Connect an order tracking service Call a remote agent How the code editor suggests methods like an IDE Compose an orchestrator from sub-agents Connect an MCP tool Duplicate and edit a Skill Escalate a hard case to an operator Create the Order return Skill Send the answer to Telegram Process data with a code node Get a strict JSON answer

Data and variables

Attach a model context resource Agent parameters and secrets Reuse settings and files with resources

Testing and debugging

What the customer sees while waiting for an operator Debug the agent step by step Find and resume past sessions Inspect traces and logs Configure quality review and see the verdict in chat A Skill unlocks in chat Test the agent in chat Time-travel the run history See the operator handoff in the run history

Triggers

What triggers are and how to start configuring them Run the agent from inbound email Run the agent on a platform event with a filter Run the agent on a schedule (cron) Connect a Telegram bot to the agent Run the agent from an HTTP webhook

API and access

First agent request through API Console Sync and async calls through API Console Embed the agent on a site Share an agent demo link

Publishing and operations

Chat settings for customers Enable the agent as an AI employee Experiment in a branch Publish and operate the agent Configure voice and limits

Quality evaluation

Build a multi-turn evaluation case First agent quality test Score an answer with an LLM judge Monitor quality in production Run tests and compare versions

Company connections

Connect external accounts Manage company MCP servers