Introduction
What is Evolvable AI?
Evolvable AI is a multi-tenant platform for building, deploying, and managing AI agents. It allows organizations to create intelligent conversational agents powered by large language models, connect them to knowledge bases and workflows, and monitor their activity in real time.
Key Concepts
| Concept | Description |
|---|---|
| Tenant | An isolated organization workspace. All data (agencies, agents, documents, guardrails) is scoped per tenant. |
| Agency | A logical grouping of agents. An agency defines the context, members, and conversation flow for a set of related agents. |
| Agent | An AI assistant configured with a model, instructions, knowledge, tools, and guardrails. Agents handle conversations with end users. |
| Guardrail | A content safety rule applied to agent inputs or outputs. Blocks or flags messages matching defined phrases, patterns, or regex. |
| Knowledge | Documents uploaded to the platform and used as context for agent responses via RAG (retrieval-augmented generation). |
| Workflow | An automated process built with a visual editor. Agents can trigger workflows as part of conversation flows. |
| Monitoring | Real-time visibility into active agent conversations and historical conversation logs. |
Main Areas
- Agencies — Create and manage agencies, assign agents, invite members and admins.
- Agents — Configure agents: model, instructions, knowledge, tools, workflows, integrations, memory, guardrails, and evaluations.
- Knowledge — Upload documents and manage memory that agents use for context.
- Guardrails — Define content safety rules and blacklist phrase groups.
- Workflows — Build and run automation workflows with a visual editor.
- Monitoring — Observe live agent sessions and browse conversation history.
- Playground Chat — Test agents directly from the admin portal.
- Demos — Embedded standalone demo applications showcasing agents in real-world scenarios.
Last updated on