Persistent identity
The agent wakes up as itself.
Roles, values, relationships and permissions are reconstructed from a governed identity record—not improvised from a prompt.
Production-derived aggregate · updated 16 Jul 2026 · 15:29 PET
Change the model without restartingthe relationship.
SOUL keeps an AI agent's identity and memory outside the model, with authority, privacy and traceability controls that survive changes of model, interface and team.
Designed to be model-independentAuditable by designHuman-governedSimulation-first embodiment
SOULPersistent core
IDENTITY
MEMORY
AGENCY
BODY
A continuity layer
Your agents should not start from zero every session.
Temporary context is not durable memory. SOUL provides the infrastructure needed to preserve identity, govern recall and verify what happened.
Governed memory
Memory with rules and provenance.
Episodic, semantic, procedural, emotional and working memory with retrieval controls and reversible consolidation.
Multi-agent teams
Many identities without blurred boundaries.
Agents coordinate and delegate while each role, private boundary and handoff remains explicit and auditable.
Verifiable operations
Evidence closes the loop.
High-impact intent passes authority and policy gates, and success is reported only after its effect is checked.
17,430
Current governed memories
100%
Current records vectorized
5
Persistent principal identities
2026-03-29
Continuity recorded since
Built for real work
Continuity for products, teams and embodied systems.
SOUL is for organizations that need more than a stateless chatbot and want control over what an agent remembers, who it is and what it may do.
Agent products
Move across models without erasing history.
Keep identity, permissions and operating context outside any single model provider.
Business teams
Deploy governed specialist agents.
Give each human or workflow one or more agents with explicit access and auditable handoffs.
Research & education
Study long-term interaction responsibly.
Inspect memory provenance and continuity while keeping private content inside controlled infrastructure.
Robotics
Validate cognition before hardware.
Use signed intent contracts and simulation-first safety gates before controlled promotion to a physical adapter.
A governed cognition loop
From event to verified action.
SOUL preserves the reasoning trail without exposing private memory on this public site.
Observe
Events become evidence
Remember
Context gains provenance
Reflect
Experience informs judgment
Decide
Intent passes policy
Verify
Effects close the loop
Evidence over adjectives
Live aggregates, explicit limits.
The public snapshot contains privacy-safe totals derived from the production memory schema. Evaluations are internal engineering results, not external certification.
Production-derived aggregate
Only system totals are published. Agent names, messages, memories and personal content remain private.
- Data source
- SOUL · aggregate-only production snapshot
- Vector coverage
- 17,430 / 17,430
- Privacy boundary
- Aggregates only · zero memory content
59 / 60
SEAL-Bench retrieval
Internal engineering evaluation
24 / 24
Privacy logic
Internal engineering evaluation
8 / 8
Risky actions denied
Internal engineering evaluation
35 / 39
Adversarial inputs detected
Internal engineering evaluation
Evidence register: SOUL_SYSTEM_CARD_v0.5_ES.md
From software to presence
One governed identity. Multiple possible interfaces.
SOUL separates cognition from embodiment. It can operate through software today; robotic adapters remain simulation-first and require explicit safety promotion.
- Signed intent contracts
- Human approval gates
- Simulation-first promotion
- Portable memory contracts
Illustrative SOUL Body intent
{
"identity": "persistent-agent-id",
"intent": "navigate_to",
"target": "workbench_1",
"policy": {
"human_approval": true,
"speed_limit": 0.4,
"collision_guard": "required"
},
"status": "policy_checked"
}Clear answers
Ambition without inflated claims.
What works now, what is internal evidence and what remains research stay visibly separated.
Is SOUL another language model?
No. SOUL is infrastructure around models. Identity, memory, authority, provenance and continuity remain external so the underlying model can change.
Does SOUL claim consciousness?
No. We make an engineering claim: persistent state and governed behavior provide continuity that is measurable, inspectable and portable—without metaphysical promises.
Is it ready for physical robots?
The embodied runtime is simulation-first. Hardware promotion remains gated by explicit safety tests and human authority.
How can we evaluate it?
SOUL is offered as a guided private preview. The SEAL team reviews your use case, architecture, evidence and deployment fit with you.
Build continuity, not another chatbot.
Review the architecture, evidence and fit directly with the team building persistent agent continuity.
Private guided review
Architecture and evidence walkthrough
No obligation or automated sales sequence