VedAGI Platform

Verified Model.

One verified decision kernel: proof-backed numbers, governed actions, your data stays on your install. Powering ClarityOps, MyCelAI, and Serlyx.

What it is

A shared decision ontology with proof, not a fourth product.

ClarityOps, MyCelAI, and Serlyx are industry overlays on one kernel. The Verified Model is that kernel: a shared object model where writes go through governed actions, security markings travel with the work, and every number that counts as a Figure is computed under proof.

It rhymes with a modern decision ontology (data, logic, actions, security) and adds what regulated work needs: code hash, sources, validators, and an audit trail you can put in front of a board.

Object model

Named types. Named accountability.

Plain language for the types that travel across every platform.

Sources

The inputs: datasets, repos, waveforms, telemetry, documents. Hashed and marked so lineage is checkable.

Runs

One execution of a pipeline, crew, or underwriting job, with status, inputs, outputs, and proof attached.

Figures vs Estimates

A Figure is a computed value with units, code hash, sources, and validators. An Estimate is labeled uncertain output. They are never silently swapped.

Findings

Structured conclusions with evidence links: what was seen, how severe, and what supports it.

Actions

The only path that mutates the store. Functions and models may propose. Actions apply, with criteria and markings.

Proof

Code blob hash, sandbox profile, validation report, and audit record. How a Figure defends itself.

Figure gate

A model cannot mint a Figure.

If a value touched an LLM context, it is an Estimate (labeled) or it does not render as a Figure. Promotion means a human Action, then recompute in a sandbox under validators. That is the hard line between narrative assistance and numbers you can defend.

Data locality

Your data stays on your install.

Local-first by default

Air-gap is a first-class path, not a special case. Cloud is an optional spoke, not the architecture.

Customer-owned record

Process data stays on the customer enterprise or install. VedAGI is not the system of record for your operational data.

Markings travel

Security markings propagate through Actions, exports, and edge sync. Discretionary filters alone are not enough for regulated output.

Models

Bring your own model.

Use a local model or your own cloud subscription. Marking-aware routes keep sensitive classes (such as PHI) on local-only paths and fail closed when a route would violate policy. Models may plan and narrate. They still cannot invent Figures.

Engagement

We work with a few enterprise and public sector teams at a time, because doing this properly means getting into the real work, not a demo.

VedAGI. Intelligence for Humanity.