⬡ Fronesis Labs · 2026

The
verification
layer for AI

Leibniz Layer is the cryptographic foundation for AI agent accountability. Every decision, every action deterministically sealed, tamper-evident, auditable.

→ Try DCL Evaluator Read the concept
⬡ SHA-256 HASH CHAIN ⬡ DETERMINISTIC COMMITMENT ⬡ EU AI ACT READY ⬡ MULTI-LLM SUPPORT ⬡ MCP PROTOCOL ⬡ OFFLINE CAPABLE ⬡ MERKLE TREE AUDIT ⬡ TAMPER-EVIDENT ⬡ SHA-256 HASH CHAIN ⬡ DETERMINISTIC COMMITMENT ⬡ EU AI ACT READY ⬡ MULTI-LLM SUPPORT ⬡ MCP PROTOCOL ⬡ OFFLINE CAPABLE ⬡ MERKLE TREE AUDIT ⬡ TAMPER-EVIDENT

⬡ What is Leibniz Layer™

A new layer
in the AI stack

Leibniz Layer™ is a cryptographic verification protocol for AI agent decisions. Named after Gottfried Wilhelm Leibniz — mathematician, logician, and pioneer of deterministic reasoning — it brings the same rigour to modern AI systems.

Just as TLS secured the web without changing HTTP, Leibniz Layer secures AI agent behaviour without changing your existing stack. It sits between your agents and the world, sealing every decision into a tamper-evident chain.

DCL Evaluator is the first product implementing this layer — available today.

05 · APPLICATION
Your AI Agent
LangGraph, AutoGen, custom pipelines
04 · VERIFICATION
⬡ Leibniz Layer™/div>
Cryptographic commitment & audit trail
03 · PROTOCOL
MCP / OpenAI API
Model Context Protocol, REST
02 · MODEL
LLM Provider
Claude, GPT-4, Grok, Gemini, Ollama
01 · COMPUTE
Infrastructure
Cloud, on-premise, air-gapped

⬡ Core properties

01
Deterministic

Identical input + policy = identical decision. 100% reproducible across runs, machines, and time. Unlike probabilistic LLM filters, Leibniz Layer never surprises you.

02
Tamper-evident

Every agent action is sealed with SHA-256 into a Merkle hash chain. Modify any past record — the entire chain invalidates. Mathematical proof of integrity.

03
Auditable

Full audit trail exportable as tamper-evident PDF, JSON, or CSV. Ready for regulators, insurers, and legal proceedings. Every decision leaves a permanent record.

04
Model-agnostic

Works with Claude, GPT-4, Grok, Gemini, DeepSeek, and local Ollama models. The verification layer is independent of which LLM powers your agent.

05
Offline-capable

Run fully air-gapped with local models. Zero data leaves your infrastructure. Critical for regulated industries: finance, healthcare, defence, government.

06
Compliance-ready

Built-in policy templates for EU AI Act, GDPR, Finance (AML), Medical (HIPAA), and Red Team scenarios. Leibniz Layer speaks the language of regulators.

⬡ First implementation

Powered by Leibniz Layer

DCL
Evaluator

The first desktop-native, offline-capable, multi-LLM audit tool built on Leibniz Layer. Connect via MCP in minutes. Available now for individuals and teams.

  • Desktop-first, works fully offline
  • MCP server endpoint: mcp.fronesislabs.com
  • 4 audit tools: commit, verify, chain, report
  • Tamper-evident PDF compliance reports
  • Drift detection with 4-level escalation
  • Source-available, patent-pending architecture

→ Get DCL Evaluator
# Connect via MCP (Claude, Cursor, VS Code) "mcpServers": { "dcl-evaluator": { "url": "https://mcp.fronesislabs.com/sse", "headers": { "x-api-key": "your-key" } } } # Available tools dcl_commit → seal agent action dcl_verify → check integrity dcl_get_chain → full audit trail dcl_report → compliance PDF # Health check https://mcp.fronesislabs.com/health

⬡ Who it is for

Developers
Build with confidence

Plug into your agent pipeline in minutes. Understand why your agent makes decisions. Stop debugging black boxes.

Enterprise
Prove compliance

Cryptographic audit trails ready for EU AI Act, GDPR, and sector regulators. Evidence your AI governance programme actually works.

Platforms
Embed the standard

License Leibniz Layer™ technology to add native verification to your platform. Give your customers guarantees without building from scratch.

Investors
The next layer

Every AI agent economy needs a verification layer. Leibniz Layer™ is that layer — patented, shipping, and growing. First mover in a mandatory market.