01 Automatic Pipeline Creating 02 Pipeline and Agents 03 Creating New Agents 04 Memory and Monitoring
Platform · 02

Pipelines connect agents.
Agents do the work.

A Revula pipeline is a graph of agents — each one responsible for a specific step. Together they execute complex, multi-step workflows that reason, branch, and adapt.

The visual canvas

Your pipeline as a live graph

Each node on the canvas is an agent. Each edge is a communication protocol. Drag to connect, click to configure, activate to run — all in the same interface.

One-directional — data flows A to B onedi
Bidirectional — agents exchange context bidi
Uses — A calls B synchronously uses
Revula · Canvas
Email Trigger entry point Extractor extraction · 3 skills Matcher reasoning · 2 skills ERP Lookup →
Pipeline structures

Pipelines can be any shape

Sequential
Agents execute in order, each passing their output to the next. Best for linear document processing, approval chains, or data transformation pipelines.
Branching
Agents route data to different downstream agents based on conditions — routing exceptions to escalation agents, flagged items to review queues, or results to different outputs.
Iterative
An agent calls another in a loop — refining output, retrying failed steps, or processing a list of items one by one. The loop stops when a condition is met.
What's inside an agent

Every agent is a configured reasoning unit

Agents aren't black boxes. Every decision they make is governed by explicit instructions, skills, and tools — all of which you can inspect and edit.

Instructions
Plain language rules that tell the agent what to do, what to avoid, and how to handle edge cases. Written by you, editable at any time.
Skills
Reusable knowledge modules — domain expertise, standard operating procedures, reference documents — injected into the agent's context at runtime.
Tools
External capabilities the agent can call — read from a database, send a Slack message, query an API. Tools are executed by the platform, not hallucinated by the LLM.
Memory
Each agent can read from and write to the pipeline's memory — persisting context across runs, learning from feedback, and building an institutional record over time.

See it in action

We'll walk you through a real pipeline built for your workflow during our pilot call.

Contact us