cognis-digital

MCP server hardening linter — capability declarations, transport, tool descriptions

LLM red-team harness — OWASP LLM Top 10 + MITRE ATLAS attack packs

Deterministic, token-budgeted, AST-aware repository context packs for LLM agents — with a hash-chained provenance row for every read. Zero-dependency CLI + MCP.

No-train, on-prem code knowledge graph served to AI agents over MCP, with a hash-chained audit row for every read.

The AI content team you own: voice-trained hooks, scripts, captions, thumbnails & multi-platform packaging. Offline by default, your model when you want it.

Verifiable AI-agent identity + multi-hop delegation chains anchored to a human principal (the unsolved 2026 agent-auth gap)

Infer where a photo was taken using a local uncensored vision + reasoning model (OSINT/geoint, 100% local)

Local model router / proxy across Ollama, vLLM, and cloud with fallback

Versioned prompt / template registry with A/B and rollbacks

Batteries-included local RAG pipeline — ingest, index, serve

Local skill registry and installer for AI agents

Runtime allowlist and policy for agent tool-calls

AI Agent Permission & Access Auditor — surfaces the lethal trifecta of credentials + injection + reach

Auto-generated NIST AI RMF / EU AI Act Annex IV model & system cards

Embedded bias probe suite — demographic / occupational / geographic

Runtime agent firewall — PII redaction, rate limits, policy enforcement

LLM hallucination & grounding auditor for RAG systems

Local LLM cost & token forensics proxy with anomaly detection

RAG corpus poisoning detector — embedding anomalies, backdoor triggers

Portable long-term memory store for agents, exposed over MCP

Lightweight MCP-native CRM pipeline with email sequences