VALORA AI WORKFLOW ORCHESTRATION

Product Presentation · 2026

AI Workflow Orchestration

Amplify your development teams with specialised AI agents that automate the complete software lifecycle.

v2.4.1Version
TypeScriptRuntime
MITLicence
Node 18+Requirements
“
The future of software development is not about replacing developers, but amplifying their capabilities through intelligent AI collaboration.
— VALORA — README.md
01

— The Context

The Challenge of
Modern Development

Why teams struggle to deliver quickly, without technical debt

Traditional Development

Constant friction

  • Constant context switching between tools
  • Manual documentation — often forgotten
  • Inconsistent code reviews depending on the reviewer
  • Time-consuming PRs — writing, formatting, tracking
  • Sloppy commit messages, unreadable history
  • Variable quality depending on team workload

With VALORA

Unified & automated workflow

  • Unified workflow in a single CLI
  • Auto-generated documentation at every stage
  • Exhaustive AI reviews, reproducible
  • PRs created in one command with full context
  • Intelligent conventional commits
  • Consistent quality, independent of workload
02

— Overview

What is
VALORA?

An acronym, a vision, a platform

Overview

Versatile Agent Logic for Orchestrated Response Architecture

Powerful CLI

TypeScript command-line interface — 25 commands covering the entire lifecycle, from spec to PR.

11 Specialised AI Agents

Each agent is an expert — lead, QA, security, design, infrastructure — coordinated by the orchestration engine.

8 Governance Phases

Rigorous lifecycle from initialisation to feedback, with quality checkpoints at every stage.

11Agents
8Phases
25Commands
3Execution modes
15MCP Servers
03

— Specialised Agents

The 11 Specialised
Agents

The right expert for every task

Specialised Agents

Every agent, a dedicated expert

@lead Technical Lead Architecture oversight, planning, comprehensive code reviews
@product-manager Product Manager Requirements, prioritisation, PRD, backlog, product feedback
@software-engineer Software Engineer Implementation specialists — frontend, backend, fullstack
@platform-engineer Platform Engineer Infrastructure, CI/CD, DevOps, deployment, monitoring
@qa QA Engineer Unit, integration, e2e tests — systematic quality assurance
@secops-engineer SecOps Engineer Security, compliance, audit, vulnerability detection
@ui-ux-designer UI/UX Designer Design, accessibility, user experience, components
@asserter Asserter Implementation validation, acceptance criteria verification
04

— Development Cycle

Complete 8-Phase
Lifecycle

From initialisation to feedback — no step missed

Development Cycle

The 8 phases of the VALORA lifecycle

01
Initialisation
Project configuration, initial context
valora init
valora config setup
02
Planning
Specs, PRD, prioritised backlog
valora refine-specs
valora create-prd
valora generate-docs
valora create-backlog
03
Context Gathering
Codebase analysis, implementation plan
valora refine-task
valora plan
valora review-plan
04
Implementation
Code execution by dynamic agents
valora implement
valora assert
05
Testing
Unit, integration, e2e suites
valora test
06
Review
Code, functional, security review
valora review-code
valora review-functional
07
Delivery
Conventional commits, automated PRs
valora create-pr
08
Feedback
Results capture, persistent memory and monitoring
valora feedback
valora consolidate
valora dash
05

— Technical Architecture

How It
Works

4-layer architecture, extensible and modular

Technical Architecture

4-layer modular architecture

CLI Layer
→
Commands
Assistant
Dashboard
Formatted output
src/cli/
Orchestrator
→
Pipeline Executor
Stage Executor
Context Manager
Session Manager
Memory System
src/executor/ · src/memory/
Agents
→
Registry
Dynamic Selection
Prompt Loading
11 Defined Agents
data/agents/
LLM Layer
→
Anthropic
OpenAI
Google
MCP Sampling
3-tier fallback
ASTCode Intelligence
LSPSymbol Index
MCPTool Protocol
SessionContext persistence

Technical Architecture · New in v2.4

Biological memory of agents

Episodic · half-life 7 d

Timestamped observations per session — detected errors, code patterns, local decisions

Semantic · half-life 30 d

Patterns consolidated by Jaccard similarity — cross-session architectural knowledge

Decisions · half-life 21 d

Persistent architectural decisions — automatically invalidated by git log

  • Exponential decay — Ebbinghaus curve, automatic pruning without intervention
  • Reinforcement on read — each access adds 2 d to the entry's half-life
  • Error amplification — errors remembered 2× longer
  • Confidence levels — verified → observed → inferred → stale
  • Zero dependencies — pure JSON, stored in .valora/memory/
terminal 
# Manual consolidation
valora consolidate

# Simulation without writing
valora consolidate --dry-run

# Auto-triggered after feedback
06

— Execution Modes

Three Execution
Modes

From zero configuration to full autonomy

Execution Modes

Choose your level of autonomy

Recommended
01
MCP Sampling
Free*
Via your Cursor subscription. No API key, no configuration required.
Uses Claude via Cursor
Zero additional cost
Available immediately
Maximum autonomy
No API Key
02
Guided Completion
Free
Generates structured prompts to copy into any AI assistant.
Works everywhere
Optimised prompts
Compatible with all LLMs
Full human control
Advanced
03
Direct API
Pay-per-use
Direct calls to Anthropic, OpenAI or Google APIs — fully autonomous execution.
Full autonomy
Maximum power
Automatic fallback
Complete call audit

Model Optimisation

The right model for every task

Model Usage %
GPT-5 Thinking Deep analysis, planning 31%
Claude Sonnet Implementation, reviews 31%
Claude Haiku Quick tasks, validation 38%

Strategic allocation

Strategic 31% Exec. 31% Fast 38% Balance Cost / Speed / Quality
07

— Security & Compliance

Enterprise
Security

Multi-layer protection for demanding environments

Security & Compliance

6 levels of protection

Credential Guard
Redaction of environment variables, blocking sensitive files, output scanning
Command Guard
Blocks exfiltration, dangerous network patterns, eval and remote access
Injection Detection
Scanning tool results with risk score, quarantine and automatic redaction
MCP Hardening
Tool definition validation, drift detection, approval workflows
Supply Chain
Frozen lockfile, installation scripts blocked, vulnerability overrides forbidden
Audit Logging
Complete operation traceability with security event tracking
08

— Extensibility

How to Extend
VALORA

Plugin architecture — everything is overridable

Extensibility

Plugin architecture — everything is extensible

Structure .valora/

Override folder
.valora/
├── config.json # project config
├── external-mcp.json # MCP config
├── hooks.json # hooks and enforcement config
├── agents/ # custom agents
├── commands/ # custom commands
├── prompts/ # custom prompts
├── templates/ # custom templates
├── sessions/ # session state
└── logs/ # logs

Custom Agents

Redefine an agent
# .valora/agents/@lead.md
---
name: @lead
model: claude-opus-4-6
focus: microservices
---

# Your custom prompt
You are a microservices architecture expert for fintech...

External MCP Servers

15 available integrations
Playwright Figma GitHub
Terraform Firebase MongoDB
Grafana Storybook Elastic

Each server activated with user approval.
Configuration in external-mcp.default.json

Quick Reference

25 commands across 3 phases

Planning · 13
refine-specsRefine specifications
create-prdGenerate PRD document
create-backlogBreak down into tasks
generate-docsGenerate technical documentation
generate-all-documentationParallel generation (~6 min faster)
fetch-taskFetch the next task
refine-taskClarify requirements
gather-knowledgeAnalyse codebase context
planCreate implementation plan
plan-architectureHigh-level architecture plan (Phase 1)
plan-implementationDetailed implementation plan (Phase 2)
validate-planPre-validation of the plan (60–70% of issues)
review-planValidate plan quality
Implementation · 8
implementExecute code changes
assertValidate implementation
testRun test suites
validate-coverageValidate test coverage
pre-checkQuality pre-checks before review
validate-parallelassert + review-code in parallel (~50% faster)
review-codeCode quality review
review-functionalFunctional review
Delivery · 4
commitCreate conventional commits
create-prGenerate the pull request
feedbackCapture results
consolidateConsolidate agent memory

Getting Started

From zero to your first PR in 6 commands

terminal 
# 1. Global installation
pnpm add -g @windagency/valora

# 2. Initialise in your project
cd my-project && valora init

# 3. Create the implementation plan
valora plan "Add OAuth authentication"

# 4. Run the implementation
valora implement

# 5. Review + conventional commit
valora review-code && valora commit

# 6. Automatic pull request
valora create-pr
Zero configuration required with your Cursor subscription
VALORA

Multi-agent orchestration in service of developers

→ 11 specialised agents automatically coordinated for every task
→ Zero configuration — works immediately with your Cursor subscription
→ Extensible by design — agents, commands and prompts overridable per project
pnpm add -g @windagency/valora

MIT · Node.js 18+ · TypeScript 5.x