Use Cases

Multi-Agent Fleets

Run dozens or hundreds of agents with isolated per-bot memory namespaces. Each bot maintains its own grounded recall, self-model, and procedural skills — cross-contamination rate stays at 0% by construction, not by convention.

What you get:

• Per-bot memory isolation with memory_namespace_* tools
• Persistent self-models that survive restarts
• Procedural skill libraries shared or per-bot
• Audit trail on every read/write

Long-Running Agents

Agents that work across days, weeks, or months need memory that persists beyond a single context window. mem20's ledger and deep-store give you that persistence — with full epistemic status tracking so the agent knows what's certain vs. speculative.

What you get:

• Working memory that survives session restarts
• Vector + BM25 hybrid recall for long-range retrieval
• Epistemic status on every record (certain / uncertain / quarantined)
• Epistemic veto prevents acting on low-trust premises

Research & Simulation

Run counterfactuals, explore hypotheticals, and model scenarios — without polluting grounded recall. mem20's simulation partition is a sandbox you can write freely into, and promotion to grounded status requires real evidence.

What you get:

• imagination_counterfactual() for "what if" exploration
• world_model_do() for causal simulation
• Prediction ledger that tracks forecasts against outcomes
• Evidence-based promotion (no honor-system upgrades)

Customer-Facing AI

Chatbots and assistants that remember user preferences, past conversations, and stated needs — while keeping those memories grounded in real interactions, not hallucinated personas.

What you get:

• User-specific memory pins (memory_pin)
• Grounded preference tracking
• Auditability for compliance
• No cross-contamination between users

Creative & 3D Workflows

With optional Blender and Unity modules, agents can persist project state, remember design decisions, and build on past creative work across sessions.

What you get:

• 7 Blender tools (modeling, rendering, scene management)
• 5 Unity tools (build, test, script generation)
• Graceful degradation when external apps are absent
• Extensible template for Figma, Unreal, and more

Safety-Critical Agents

For any deployment where an agent acting on false memory is unacceptable. The contamination firewall, epistemic veto, and audit engine make mem20 the memory layer you can actually trust in production.