Sovereign AI for Everyone!
Built local-first, security-native, owned by you.
Your memory, your agents, your rules.
Why this exists
Sovereign memory for local-first AI
AI assistants are becoming memory systems: places where we ask questions, store context, make decisions, and build up a working model of who we are. That memory should not be trapped inside a vendor account or disappear when a subscription ends.
LegionForge is built around a simpler premise: your memory belongs to you. You should control what gets inserted, updated, deduplicated, enhanced, forgotten, or resolved when facts conflict. You should be able to inspect and move your data, tune your agents, and decide which tools can touch which parts of your identity.
LegionForge provides a user-first, security-conscious, open-source path for local AI: owned by you, running under your rules, with deterministic validation where it matters and LLMs used where they actually help.
Five non-negotiables
Design principles
Sovereignty first
Your memory, agents, rules, and data stay under your control. Inspect, tune, update, move, or forget any part of the system.
Local-first custody
Core state lives on infrastructure you control. Cloud services are optional dependencies, not the source of truth.
Human authority
Irreversible changes, sensitive writes, and memory conflicts cross a human-controlled boundary by default.
Deterministic guardrails
Rules, signatures, trust scores, and capability checks run before the model gets a vote.
Right to exit
Your memories and audit trails should be portable. Leaving a vendor should not mean losing yourself.
The ecosystem
Build with what fits
LegionForge is a family of independent, composable projects for sovereign AI: memory governance, local agents, voice interfaces, orchestration, diagnostics, and security tooling. Use one piece, or wire the stack end-to-end.
LegionForge Framework
Local-first AI agent framework, security in the foundation
coreGuardian
Deterministic security sidecar for any agent framework
appJeli
Security and governance layer for sovereign personal memory
appADHD-OS
Personal assistant framework for those with ADHD
appConvoBox
Local, backend-agnostic voice frontend for CLI coding agents
orchestrationBriarios
Security-native orchestration meta-layer for parallel agent workstreams
researchIntelligence Delivery Network
CDN-style routing model for LLM workloads
toolidn-analyzer
Offline prompt profiler for LLM routing decisions
toolllm-valet
Auto-pause/resume Ollama based on system pressure
toolmcp-probe
Connectivity and configuration advisor for your MCP services
toolheadroom
System stability monitor with AI-powered diagnostics
toolhermes-tool-test-suite
pytest harness for AI agent tool-calling reliability
tooldev-rig
Shared CI workflows and audit harness for LegionForge projects
Security posture
Security in service of ownership
Security is not only about blocking bad tool calls. It is about making sure your memory, identity, and agent history remain inspectable, portable, and governed by you.
LegionForge favors deterministic checks, scoped capabilities, signed provenance, and human-controlled mutation points. The model can help reason, but it should not be the authority deciding what your system is allowed to remember or do.