QGI Logo QGI
Enterprise Runtime · Orchestration

Neural Symbolic Agents

Enterprise agent runtime

The top layer of the QGI stack. Neural Symbolic Agents orchestrate multi-agent workflows where every agent reasons through the QAG Engine — so the entire workflow has one inspectable reasoning graph, not a pile of independent black boxes.

Availability

Neural Symbolic Agents ships as a custom enterprise engagement. There is no self-serve tier: every deployment is scoped to a regulated workflow, goes through a structured pilot, and graduates into production with a named audit and governance contract.

Request an engagement →
What the runtime does

Multiple agents. One reasoning graph.

Runtime

Multi-agent orchestration on a deterministic substrate

Every agent in the runtime reasons through the QAG Engine. Hand-offs between agents propagate the same seven HSC signals, so the overall workflow has a single, inspectable reasoning graph — not a collection of independent black boxes calling each other over HTTP.

Memory

Persistent quantum-structured memory

Agents share quantum-graph memory through QGM (or the current interim memory engine before QGM GA). Success-vs-failure structure is preserved across sessions. When an investigation resumes six weeks later, the second analyst inherits the reasoning trail — not a summary.

Coordination

Dependency and conflict coordination

When two agents reach conflicting conclusions, the runtime surfaces the conflict back up — it does not paper over it with a majority vote. Dependencies between decisions are tracked explicitly so compliance can audit who-decided-what-and-why without a forensic reconstruction.

Integrations

Tools and integrations

Agents can call into the enterprise's existing tools — case-management systems, policy stores, MCP servers, NVIDIA AI Blueprints wired through the Enterprise Factory. Every call is logged as part of the reasoning graph so tool calls are as auditable as model outputs.

Deployment

Enterprise deployment and governance

Deployed as a custom engagement — VPC, on-premise, or managed. Governance artifacts (model cards, signal-level dashboards, replay tooling, audit exports) ship with the runtime. Every deployment retains QGI's audit contract.

Engagement path

From first call to audited production

Step 1

Scoping

Define the regulated workflow, success metrics, and the regulatory lens the deployment will pass under.

Step 2

Pilot

Structured pilot in a scoped workflow, with QAG Engine preview access, Q-Prime encoding, and agent runtime.

Step 3

Audit

Governance review with model cards, signal-level dashboards, and replay tooling. Signable by compliance.

Step 4

Production

Graduate into production with a named audit and governance contract. Every decision replayable on regulator request.

Frequently asked

Questions enterprise teams ask before a pilot.

What are Neural Symbolic Agents?
Neural Symbolic Agents is the top layer of the QGI deterministic stack: an enterprise agent runtime where every agent reasons through the QAG Engine. Multi-agent hand-offs propagate the same seven Hilbert-Space Compacting signals, so the entire workflow has one inspectable reasoning graph — not a collection of independent black boxes calling each other over HTTP.
How is this different from LangChain, CrewAI, or a standard multi-agent framework?
LangChain-style frameworks orchestrate probabilistic calls to foundation models. When two agents reach conflicting conclusions, the framework papers over it with a majority vote, a re-prompt, or a silent failure. Neural Symbolic Agents surfaces the conflict as an explicit HSC signal, tracks dependencies between decisions, and preserves the reasoning graph across agents. The substrate is deterministic; the orchestration is auditable.
Is Neural Symbolic Agents self-serve?
No. Neural Symbolic Agents ships as a custom enterprise engagement. There is no free tier, no self-serve API key, no starter notebook. Every deployment is scoped to a regulated workflow, goes through a structured pilot (Scoping → Pilot → Audit → Production), and graduates into production with a named audit and governance contract. We run a small number of these per quarter.
What memory does the runtime use across sessions?
Agents share quantum-graph memory through QGM (Layer 2). When QGM is not yet generally available for a specific workflow, the runtime uses an interim memory engine with the same audit properties — provenance preserved end-to-end, replay supported, and the success-vs-failure structure carried across sessions. The second analyst to pick up an investigation six weeks later inherits the reasoning trail, not a summary.
What deployment models are supported?
Managed (QGI cloud), VPC (customer cloud, data never leaves the tenant), or On-premise (air-gapped or FedRAMP-adjacent environments). Governance artifacts — model cards, signal-level dashboards, replay tooling, audit exports — ship with every deployment, regardless of model. See the Trust page for the full deployment and compliance posture.
Who is the right customer for Neural Symbolic Agents?
Regulated teams with high-stakes multi-step workflows: credit, claims, AML/KYC, compliance, insurance, capital markets, and analogous government and healthcare workflows. The common signal is a decision chain where a regulator, auditor, or investigator will eventually ask "who decided what, why, and when" — and the answer has to be defensible at the level of the individual decision, not the aggregate model.

Bring us your highest-stakes workflow.

We run a small number of enterprise pilots per quarter in regulated industries — credit, claims, compliance, AML/KYC, capital markets, and analogous government and healthcare workflows.

Request an engagement → Review our trust posture
Partner with QGI