ScaiCore: A new paradigm for agentic AI.

ScaiCore is an AI-native agent programming language featuring a rigidity spectrum that allows developers to define how strictly or flexibly agents interpret and execute their instructions.

Overview

A language designed for AI agents

Building AI agents today means cobbling together prompts, API calls, tool integrations, and orchestration logic across multiple languages and frameworks. ScaiCore replaces that with a single, purpose-built language where agents are first-class citizens.

A ScaiCore agent isn’t just a prompt - it’s a complete definition: identity, personality, model preferences, tools, memory structure, behavioural rules, and the flows that define what it does. All in one place, versioned, testable, and deployable through ScaiGrid.

The Language

Expressive, readable, powerful

ScaiCore.core

 1@core CustomerSupport {
 2  @identity {
 3    name = "Support Assistant"
 4    version = "1.2.0"
 5  }
 6  @personality {
 7    tone = "friendly, professional"
 8    rigidity = guided   // the magic spectrum
 9  }
10  @tools {
11    use knowledge_base
12    use ticket_system
13    use escalation
14  }
15  @flow handle_inquiry {
16    retrieve context from knowledge_base
17    if confidence > 0.8 { respond }
18    else { escalate to human }
19  }
20}

Everything in one definition

A ScaiCore agent is not just a prompt. It is a complete definition: identity, personality, model preferences, tools, memory, behavioural rules, and the flows that define what it does.

@identity Name, version, and metadata for deployment and registry
@personality Tone, rigidity level, and behavioural constraints
@tools External capabilities the agent can invoke
@flow Step-by-step logic with branching, loops, and human-in-the-loop

All versioned, testable, and deployable through ScaiGrid.

Core Concept

The rigidity spectrum

The most important decision in agent design: how strictly should the agent follow instructions?

Strict

Agent follows instructions exactly. No deviation, no improvisation. Ideal for compliance-critical workflows.

Guided

Agent follows the general approach but can adapt details. Good for customer service and knowledge retrieval.

Adaptive

Agent uses instructions as guidelines but makes independent decisions. Suited for research and creative tasks.

Autonomous

Agent operates independently within broad constraints. For monitoring, alerting, and self-healing systems.

Platform

The ScaiCore Console

Deployment management

Deploy agents to ScaiGrid with version control, rollback, and A/B testing.

Execution monitoring

Watch agents run in real time. See every decision, tool call, and response.

Approval centre

Human-in-the-loop checkpoints where agents pause for review before taking action.

Debugging tools

Step through agent execution, inspect memory state, and replay past runs.

Analytics

Track performance, cost, resolution rates, and user satisfaction per agent.

Agent registry

Browse, search, and discover published agents. Fork and customise existing agents.

Use Cases

What you can build

Customer support agent

Handles inquiries, retrieves knowledge, creates tickets, and escalates to humans when needed - all in a single agent definition.

Document processing

Reads documents from ScaiDrive, extracts structured data, validates against business rules, and updates downstream systems.

Workflow automation

Orchestrates multi-step business processes with conditional logic, parallel execution, and human approval gates.

Compliance monitoring

Watches data streams for policy violations, generates reports, and alerts the right people automatically.

Ready to build intelligent agents?

Explore the ScaiCore language or get in touch to discuss your use cases.

Get in Touch →