Getting Started
Key Concepts
Core terminology used across Amantra: tenants, workflows, activities, agents, and more.
Understanding these core concepts will help you get the most out of Amantra.
Tenant
A Tenant is an isolated workspace within Amantra. Each tenant has its own users, workflows, agents, data, and configuration. This enables multiple teams or organizations to use the same platform without any data overlap.
Workflow (Synapse)
A Synapse is the fundamental automation unit in Amantra. It represents a complete business process, built as a connected sequence of steps (Activities) that run in a defined order — from the moment something triggers the process to the moment it finishes.
Think of a Synapse as a flowchart brought to life: every box becomes a step that executes, and every arrow becomes a rule for what happens next.
Example
A purchase order approval Synapse might:
Receive a new PO via email
Extract data from the attached PDF using ReadAI
Look up the vendor in the ERP system
Route the request for manager approval if the amount exceeds a threshold
Send a confirmation email
Key characteristics
Deterministic execution — given the same input, a Synapse follows the same path every time (unless a condition branches it elsewhere). This predictability is what makes automated processes auditable.
Composability — a Synapse can call another Synapse as a sub-workflow, so large processes can be broken into smaller, reusable pieces instead of one giant flow.
Multiple entry points — the same Synapse can be started by different Triggers (schedule, webhook, email, event, or manual), so one process definition can serve several use cases.
Full traceability — every run of a Synapse is recorded as an Execution, with a step-by-step activity log.
Activity
An Activity is the smallest working piece of a Synapse — a single step that performs one clear function before control passes on to the next step in the sequence.
Activities are not limited to simple system calls. Amantra supports several categories of Activity, each suited to a different kind of task:
System Action — Executes a concrete operation — calling an API, updating a database, sending an email or notification
Switch — Evaluates a condition and branches the workflow down one of multiple paths
Human-in-the-loop — Pauses execution and waits for a person to give input or approval
Sub-workflow — Calls another Synapse as a nested step, keeping the design modular
AI Tool / Agent Call — Invokes a digital agent or AI tool (such as ReadAI) to handle tasks needing judgment or unstructured data
Activities are where the actual business logic lives. Choosing the right mix — a strict system action here, a human approval gate there, an AI agent call elsewhere — is what separates a rigid automation from one that can handle real-world variability such as missing data or ambiguous documents.
Each Activity’s inputs, outputs, and status are individually logged as part of the parent Synapse’s Execution record, which is what gives step-by-step visibility when something needs debugging.
Digital Agent (Agentix)
A Digital Agent is an AI-powered software robot that can autonomously perform tasks on your behalf. Unlike a workflow (which follows a fixed path), a digital agent can reason, plan, and adapt to dynamic situations.
Digital agents can be grouped into Agent Pools so the platform automatically distributes work across available agents.
Trigger
A Trigger is what starts a workflow. Amantra supports:
Schedule — Run at a fixed time (daily, weekly)
Webhook — Run when an external system sends an HTTP request
Email — Run when an email is received matching criteria
Event — Run when an internal platform event occurs
Manual — Manually launched by a user
Execution
An Execution is a single run of a workflow or agent. Each execution has:
A unique ID
Start and end timestamps
Status (Running, Completed, Failed, Queued)
Full step-by-step activity log
ReadAI (IDP)
Intelligent Document Processing (IDP) — branded as ReadAI — uses machine learning to automatically extract structured data from unstructured documents such as invoices, contracts, forms, and IDs.
MCP Orbit
MCP (Model Context Protocol) Orbit is Amantra’s integration hub for connecting AI models and external tools using the open MCP standard. It allows your digital agents to discover and use external tools dynamically.
Connections
A Connection is a configured link between Amantra and an external service or system — for example, an AI provider such as OpenAI or Mistral, a vector database such as Pinecone, a data store such as MySQL, a Google account, or an SMTP mail server. Once a connection is created, it can be referenced by name across workflows, agents, and AI tools, so credentials and configuration are managed and secured in a single place.
Models
Model management for AI-powered automations, including configuration of AI providers and models used across workflows.
Credential Vault
The Credential Vault stores all sensitive credentials (usernames, passwords, API keys, tokens) used by your automations. Credentials are encrypted and referenced by name — your automation code never contains raw credentials.
Environment
An Environment is a named configuration profile (e.g., Development, Staging, Production) with its own set of variables. You can promote workflows from one environment to another without changing the workflow logic.
Queue
A Queue manages the order and priority of work items waiting to be processed by a digital agent or workflow. Queues ensure that your automation backlog is processed reliably and in the correct order.
ITSM
IT Service Management (ITSM) within Amantra handles:
Incidents — Unexpected disruptions that need investigation and resolution.
Service Requests — Standard requests for IT services (e.g., new user accounts, software).
Tasks — Work items assigned to agents or users.