Skip to Content
Welcome to the new Evolvable AI docs! 👋
User GuideIntroduction

Introduction

What is Evolvable.ai?

Evolvable.ai lets you build, deploy, and manage AI-powered assistants — called agents — that can hold conversations with users, answer questions from your own documents, run automated processes, and connect to external services.

Everything is organized around three core capabilities:

  • Agencies & Agents — create AI assistants, configure their behavior and capabilities, and group them under agencies that manage access and membership.
  • Knowledge Base — upload your documents so agents can search them and give accurate, grounded answers using retrieval-augmented generation (RAG) and knowledge graph traversal (GraphRAG).
  • Workflows — automate multi-step processes by connecting actions together in a visual canvas, triggered manually, on a schedule, or by external events.

Core Concepts

Agencies

An agency is the top-level organizational unit — a workspace that holds a set of agents and the users who interact with them. Each agency has admins who manage configuration and members (users) who can chat with the agents inside it.

Agents

An agent is an AI assistant you configure and deploy. You choose its language model, write its instructions, and decide what capabilities it has:

  • Tools — let the agent fetch data or take actions during a conversation
  • Workflows — trigger automated processes in response to a conversation
  • Knowledge retrieval — let the agent search your documents before answering
  • Guardrails — constrain what the agent can say or do

Agents can also be chained together: a front-facing agent can hand off conversations to specialized internal agents behind the scenes.

Knowledge Base

The Knowledge Base is where you upload documents that agents can search. When a user asks a question, the agent retrieves the most relevant content and uses it to generate a more accurate, specific answer. Two retrieval methods are available:

  • RAG — finds the most semantically similar passages using vector embeddings
  • GraphRAG — extracts entities and relationships from your documents and traverses them to answer complex, multi-hop questions

Documents are organized with tags, which also control which agents can access which content.

Workflows

A workflow is a sequence of automated steps that run in response to a trigger. Each step is a node — a discrete action such as sending an email, calling an external API, running a database query, or invoking an AI agent. Nodes are connected on a visual canvas to define the order of execution.

Workflows can be triggered:

  • Manually — on demand by a user
  • On a schedule — using a cron expression
  • By a webhook — when an external system sends a request

How They Work Together

A typical setup might look like this:

  1. You create an agency and invite team members.
  2. You upload product documentation to the Knowledge Base and tag it.
  3. You create an agent, enable RAG, and point it at those tags so it can answer product questions.
  4. You attach a workflow to the agent that, when triggered mid-conversation, queries a database and sends a follow-up email.
  5. Users interact with the agent through chat and get accurate, context-aware answers backed by your own data.

Where to Go Next

Last updated on
Agencies