The Indwel platform
Indwel supplies the whole application substrate: source collection, governed connectors, evidence, cognition, modular inference, agents and actions, application runtime, Platform API, memory, settlement, receipts, and deployment control.
The architectural divide
OpenAI / ChatGPT · Anthropic / Claude · Microsoft Copilot · Google Gemini · xAI / Grok · LangChain and agent platforms
These systems add tools, context, agents, approvals, security, and observability around a model-centred product.
Indwel moves the objective, rules, evidence, deterministic control, authority, action, and durable result outside that centre.
The inference boundary
Models contribute probabilistic inference where ambiguity, synthesis, classification, explanation, or judgement require it.
They do not own the objective, the rules, the evidence, the authority to act, or the result that becomes durable.
Choose the model. Keep the authority.
The cognitive advantage
Frontier assistants optimise individual responses. Indwel optimises the complete cognitive act across objective fidelity, evidence, computation, inference, verification, correction, and continuity.
The same leading models can be selected, compared, and replaced inside that architecture. Their intelligence is no longer asked to carry the entire process alone.
Cognition should not be left to chance.
What comes with Indwel
These are not isolated features or a menu of future integrations. They are the cooperating layers of one operating system, available to applications through the same authority, evidence, inference, action, and receipt model.
Pluggable collection infrastructure for local and cloud files, structured data, semantic units, retrieval, source lifecycle, provenance, custody, currentness, contradiction handling, and evidence admission.
Files, databases, SaaS systems, institutional APIs, cloud storage, and governed current-information survey enter through explicit source, permission, custody, and freshness boundaries.
A governed cognition kernel coordinates objective, planning, acquisition, evidence, reasoning, verification, correction, memory, and continuity. Models remain pluggable providers behind a controlled inference boundary.
Agents can plan, use tools, gather information, propose effects, and continue work—but action policy, approval, review, reversibility, and final authority remain outside the agent.
Versioned contracts expose apps, workflows, cases, runs, schemas, policies, evidence, actions, explainability, artifacts, and receipts without leaking internal kernel machinery.
Verified continuity is separated from chat history. Final results, unresolved matters, evidence, actions, corrections, and obligations become durable and inspectable.
Identity, roles, tenancy, security, model-provider administration, monitoring, cost control, deployment proof, rollback, and environment authority complete the system.
What it makes possible
Deployment authority
The governing model remains intact across deployment postures. Identity, tenancy, data custody, model providers, network boundaries, operations, and contractual controls are configured for the institution—not delegated to the model.
A managed operating environment for teams that need the complete platform without owning its underlying infrastructure.
A dedicated environment with institution-specific identity, networking, data, providers, observability, and deployment authority.
A licensed posture for organisations that require deeper operational custody, controlled integration, or specialised deployment terms.
Specific service levels, security commitments, data handling, regulatory obligations, and deployment controls are established only in the applicable written agreement.
Test the complete system