The operating model for the agent-first organization.
The twentieth-century organization ran on people moving work through departmental software. The organization taking its place runs on agents that do the work, with people holding the authority. General Reasoning builds its operating model: every actor attributed, every boundary enforced, every action on the record.
It is one stack, built to the CRC isolation standard and on open standards we publish, and deployed on dedicated infrastructure up to hardware you control. We build it first for the organizations with the most to prove: financial services, healthcare, energy, defense and mining.
An operating model, not a bundle of apps
SaaS gave every department its own application, its own users and its own logs. That model assumed a person at every keyboard. Agents break the assumption, and bolting them onto the old estate multiplies the edges nobody governs. The agent-first organization needs a single operating model that every agent and every person acts within.
| The SaaS-era organization | The agent-first organization | |
|---|---|---|
| Who does the work | People, clicking through each vendor's screens | Agents, with people holding and delegating authority |
| Where control lives | Inside each application, configured separately | In one authority model that spans the organization |
| How evidence is made | Logs pieced together after something goes wrong | A record written as each action happens |
| How software is acquired | Bought per function, one silo at a time | Generated from a specification on a shared substrate |
| What limits speed | Headcount and integration projects | The risk you can formally prove and accept |
One stack, greater than its parts
Enterprise applications run inside an enterprise reasoning engine, behind a protected gateway, with one authority model and one record beneath them all. Each layer is useful alone. Together they protect every graph edge an agent can reach, out of the box.
Three questions every agent action has to answer
A regulator, an auditor or your own incident team will ask the same three things. The stack answers each one at the moment the action happens, instead of reconstructing it afterward.
Who allowed this?
Every principal, human or agent, acts under an attributed grant. Risk that is accepted is accepted formally, by a named owner.
Where was it permitted to go?
Isolation surfaces and traversal limits decide which edges an agent may cross. Reachability alone never grants authority.
What actually happened?
Unattested data is a breach waiting to happen at agent speed. Every change is recorded, and the evidence behind each conclusion stays checkable.
Agents fail at the edges between systems
An agent is useful because it acts across systems. Every hop between systems that don't share an identity model or a record is an edge nobody governs. Monitoring tools can observe some of that traffic. Only a shared substrate can attest to it.
A typical SaaS estate
- Each vendor has its own identity model, logs and clock.
- Agents hop between platforms carrying tokens, not authority.
- Incidents are rebuilt by stitching together logs that don't match.
- Agent autonomy is capped by the least-governed hop.
The General Reasoning stack
- One authority model across every workflow and agent.
- Every edge carries its grant and attribution when it is crossed.
- Each organization and surface keeps its own isolated record. Coherent, not monolithic.
- Autonomy rises to match the risk you can formally accept.
You don't have to replace everything to start. Move the regulated workflows onto the stack first, and keep what works while the rest follows.
What you run
Each product is one part of the same operating model, sharing its authority, boundaries and record. Each has its own site with full technical detail. Status is stated plainly: we would rather you trust the badge than discover it.
DXMachine
In developmentThe applications people and agents work in, each one generated from a specification and governed from the first record. The first target is agile portfolio and backlog management for regulated delivery teams.
dxmachine.com →Chandra Enterprise
BetaGoverned records, forms and workflow on a per-organization Chandra record. Running today in client beta deployments.
chandrahub.net →Chandra Passport
In developmentAttributes every principal, human or agent, and governs what each may do: access requests, role bonding and agent capability badges, with separation of duties and every denial recorded.
AletheiaGraph
In qualificationAn enterprise database and reasoning engine in one. It keeps stored, inferred and proposed state distinct, and records why each answer can be believed.
aletheiagraph.com →AegisGenera
In developmentA hardened gateway in front of any compute backend. It enforces policy on every request crossing into or out of the governed estate.
aegisgenera.com →Industry Configurator
LiveMaps your regulatory domain onto a governed structure in minutes. More than 22 regulated domains are covered today.
config.genreason.com →Software is getting cheap. Governance isn't.
SaaS earned its price because building software was expensive. That is changing fast. When a governed application can be generated from a specification, features stop being a moat. What stays scarce is the ground applications run on: authority, boundary and record.
The App Configurator is where every DXMachine application begins. A wizard produces the specification; nobody hand-writes it. The result is not frozen: every mechanism that built the application stays available inside it for its lifetime, and forms and fields are versioned, so an upgrade never needs a data migration. The first application we are generating is our own: Chandra Passport.
Built for regulated work
The Industry Configurator ships with domain vocabularies and structures for the frameworks our customers answer to.
Pricing you can model
Three line items. You pay for the platform, for each workflow we define with you, and for the engagements actually running through it.
Platform fee
The governed stack, deployed on dedicated infrastructure or hardware you control.
Workflow definition
A one-time fee for each regulated workflow configured for your organization.
Active engagements
A per-engagement fee that scales with the work you run, not with headcount.
Every licensee gets full source access to the code they run, including code generated for their applications, to examine, audit and modify. Write to us for a quote for your deployment.
Start with your industry.
Run the Industry Configurator for your domain. Regulated organizations can also join the DXMachine development advisory program and shape the first applications before general release.