F-01
Agent inheritance
Agent hierarchies with inheritance of configurations, behaviours and capabilities — maintainable, scalable systems.
|b| Open source
Our frameworks,
published in the clear.
Three open source projects ground our technical doctrine: programmable governance, persona-centric agents, sovereign architectures. Publicly verifiable on GitHub.
P-01 · Paracle
Sovereign multi-agent platform
Open source framework to build production-ready multi-agent AI applications: modular architecture, 14+ LLM providers, portable skills, enterprise-grade security compliance.
14+
LLM providers
700+
Tests
95/100
Security score
Apache-2.0
License
F-02
Native multi-provider
OpenAI, Anthropic, Google AI, xAI, DeepSeek, Groq, Mistral, Ollama, LM Studio. No lock-in.
F-03
Portable skills
Define your skills once, deploy everywhere: Copilot, Cursor, Claude Code, Codex.
F-04
Native MCP protocol
Model Context Protocol for standardised tool discovery and interoperability.
F-05
Workflow orchestration
Sequential, parallel, conditional execution. Dependency management and error recovery.
F-06
Security & compliance
ISO 27001, ISO 42001, SOC 2 Type II, OWASP Top 10, GDPR. Daily automated scans.
Apache-2.0· Python 3.10+· FastAPI· SQLAlchemy· Redis· Docker· Next.js· PostgreSQL
P-02 · PCL
Persona Control Language
Language dedicated to AI persona governance — Terraform + OPA applied to agent behaviour. Define, govern, orchestrate and audit personas in a programmable, portable, predictable way.
5,720
Tests
8
AI providers
78/100
Prod. score
Apache-2.0
License
F-01
Language + compiler
Declarative syntax, semantic analysis, multi-target codegen (TypeScript, Python, JSON, YAML).
F-02
LSP & VS Code extension
Integrated Language Server Protocol: IntelliSense, diagnostics, navigation, autocompletion.
F-03
Team orchestration
Merge strategies (Debate, Consensus), configurable quorum, sequential/parallel workflows.
F-04
8 providers supported
Claude, GPT, Gemini, DeepSeek, Ollama, Azure OpenAI, AWS Bedrock, Mock — cross-provider.
F-05
Multi-backend registry
Memory, JSON, SQLite, PostgreSQL. MCP integration to expose personas as services.
F-06
Compliance & governance
ISO 27001, ISO 42001, OWASP LLM Top 10, EU AI Act, GDPR. Compatible with agentskills.io.
Apache-2.0· TypeScript 5.3+· PCL compiler· LSP· VS Code ext.· MCP· 8 AI providers
P-03 · ARAL
Agent Reference Architecture Layers
Open 7-layer standard where persona isn't a configuration — it IS the agent. Cryptographic provenance, dynamic persona hot-swapping, compliance with 15+ international standards.
249
Tests
93%
Coverage
220+
Requirements
15+
Standards
F-01
7-layer architecture
Runtime (L1), Memory (L2), Capabilities (L3), Reasoning (L4), Persona (L5), Orchestration (L6), Protocol (L7).
F-02
Persona = Agent
Persona embodies identity. Ed25519 identity for verifiable provenance of every action.
F-03
Dynamic hot-swapping
Identity transformation without restart. 6 modes: blend, chain, debate, consensus, override, parallel.
F-04
Multi-LLM routing
9 routing strategies (cost, quality, latency, consensus) and 6 aggregation methods (best-of-n, weighted, ensemble).
F-05
Native GDPR
Data subject rights, 72-hour breach notification, privacy by design.
F-06
15+ standards
ISO 42001, ISO 23894, NIST AI RMF, EU AI Act, OWASP LLM, OpenAI Model Spec, Constitutional AI, PCL v2.1.
CC-BY-4.0 + Apache-2.0· Python SDK (L1-L7)· JSON Schema· TypeScript ref.· Go ref.· Apache 2.0 + CC BY 4.0
|b| Contribute
Want to contribute?
Our projects are open to the community. Join us to build the future of sovereign, auditable AI.
|b| Further reading
Related resources
Deep analysis
Next.js 15 & React 19 features
A technical analysis of the latest Next.js and React features relevant to production applications.
Guides & resources
Technical documentation
Engineering articles, integration guides and technical resources for Paracle, PCL and ARAL.
Contracted operations
Support & service levels
Dedicated operation contracts for platforms integrating our open source building blocks in production.