FruxonDocs

Agents

How agents work in Fruxon — anatomy, lifecycle, and the building blocks

An agent in Fruxon is a configured AI workflow with everything attached: prompts, steps, tools, knowledge, AI providers, memory, and revision history. Agents range from a single LLM call to multi-step orchestrators that branch, loop, call sub-agents, and escalate to humans.

Anatomy of an agent

PieceWhat it is
FlowThe graph of steps you build in Studio.
ParametersTyped inputs the agent accepts at run time.
AI ProvidersWhich LLMs the steps use. Mix freely across steps.
ToolsWhat the agent can do — call APIs, query DBs, run code, hand off to sub-agents.
Knowledge BaseDocuments and data sources retrieved at runtime.
MemoryPersistent facts the agent reads and writes.
ConnectorsChannels (Slack, Teams, etc.) that can deliver the agent.
RevisionsImmutable snapshots — every save is recorded.
BudgetsMonthly cost caps and alert thresholds.

The agent lifecycle

  1. Create — name, type, description.
  2. Build — design the flow in Studio.
  3. Equip — wire AI providers, tools, knowledge, memory.
  4. Test — run interactively with sample inputs (Testing).
  5. Evaluate — score candidate revisions against a golden dataset (Evaluations).
  6. Deploy — promote a revision to production (Versioning).
  7. Connect — add API callers, connectors, or sub-agent callers.
  8. Operate — watch traces, review conversations, manage costs.

Agent types

The type field is a label that helps organize agents and provides sensible defaults — it doesn't restrict capability. Pick whichever fits.

  • Chat — multi-turn conversational agents (support, Q&A, copilots).
  • Summarization — turn long content into briefings or digests.
  • Recommendation — score, rank, or suggest based on inputs.
  • Analyzer — process data and generate insights.
  • Other — anything else.

Any type can have any tools, any flow shape, and any number of steps. Type is metadata, not a runtime gate.

What's in this section

On this page