Morentix Helm
Morentix Helm is our governed agentic platform: specialist AI agents do the work, and a person approves every decision that matters through Helm Gates. We’re not offering it to clients yet.
- YouIntakeYour goal, limits and documents
- AgentsPlanSteps, risks and what done means
- A personGateA person approves the plan
- AgentsBuildWork inside the approved plan
- AgentsReviewChecked against the plan and tests
- A personGateA person approves the result
- YouDeliverHandover with the full record
Follow one request through Helm
Pick a step to see what happens there, who is involved, what a person approves and which parts of Helm are at work.
Intake
Step 1 of 7A request will come in with its goal, its limits and the documents that matter. Nothing starts yet.
- Who is involved
- The person making the request, and a planning agent that reads it.
- What the person approves
- Nothing yet. The first approval comes at the plan gate.
- What is recorded
- The request as it was submitted, and who submitted it.
- Goal
- A customer portal for order status
- Limits
- No mobile app in this request
- Attached
- Current order spreadsheet
- Requested by
- Your operations lead
In the Helm design at this step
- Helm ContextPlannedEach agent will be given your documents, your rules and the task, and only what that task needs.
- ConnectionsPlannedFiles and data will be able to come from tools you connect, authorised in your own browser.
How Helm is designed to govern AI agents
Helm Gates lead the design: AI agents do the work, and a person approves every decision that matters. Every other part is designed to work inside them.
Helm Gates
Helm is designed around human approval. It will stop for a person before work starts and again before anything is delivered, and any high-risk action in a plan will wait for a person too. Proposed improvements to how it works will never be applied without approval. Every decision and approval will be recorded, with who decided and when.
What this means for you
- Nothing will start on your work until a person has approved the plan.
- Nothing will reach you, or your customers, without a person’s approval.
- You will be able to see who approved what, and when.
Plan ready for approval
Scope: sign-in, order status, invoices. Not included: mobile app.
- Plan: customer portalApproved by your operations lead
- High-risk step: import of old ordersWaiting for a person
- Delivery: first versionNot reached yet
Every decision and approval recorded
Helm Agents
Helm Agents are designed to be specialists for planning, building, reviewing and validating work, each with a defined job. They will work inside the limits of an approved plan and stop at each gate for a person. An agent will not approve its own work: review and validation are separate jobs. Who does what will be set out in the plan before work starts.
What this means for you
- Each piece of work will be done by an agent with one clear job.
- Work will be checked by a different agent from the one that produced it.
- A person will still approve the plan and the result.
- 1Planning agentDrafts the plan, risks and scope
- 2Building agentDoes the work the plan describes
- 3Reviewing agentChecks the work against the plan
- 4Validating agentRuns the tests before the gate
- A person approves at each gate
Helm Memory
Helm Memory is designed to keep what each engagement teaches: what worked, what to avoid and what a client prefers. Lessons will be scored for quality, and weak ones will go to a person for review instead of being kept automatically. Planning will draw on that memory, so each project can start from what earlier ones learned. A client’s memory will stay inside that client’s own deployment.
What this means for you
- Later projects will start from what earlier ones learned.
- Lessons will be reviewed before they are kept.
- Your knowledge will stay in your deployment, never shared with another client.
Lessons kept, after review
- Worked: weekly demos of each increment
- Avoid: importing old orders without a check
- Client prefers: invoices as PDF
Helm Context
Planned: Helm Context is a way to give each agent the right company and client knowledge for its task, and only that. It will assemble what an agent is told from your documents, your rules and the task in hand. It is also designed to control what each agent is allowed to see. Anything arriving from outside tools will be translated into the same format before an agent sees it.
What this means for you
- Agents will work from your own knowledge rather than guesses.
- Each agent will see only what its task needs.
- Your rules will travel with every task.
- Company knowledge: how we work
- Client knowledge: your rules
- The task: invoice downloads
Building agent
Given only what invoice downloads need; not the client’s payroll files.
Connections
Planned: Helm is designed to connect to your tools through OAuth and MCP. You will authorise each connection in your own browser, and credentials will stay in your own deployment. Every action that writes, sends or spends will wait for a person’s approval first. Reading from a tool and acting in it will be treated differently.
What this means for you
- Agents will be able to work with the tools you already use.
- Nothing will be sent, changed or paid without a person’s approval.
- Your credentials will stay with you.
Connected through OAuth and MCP · reading is allowed, acting needs approval
Works across AI providers
Helm is designed to work across many AI model providers, including open-weight models, so no single vendor is load-bearing. Each task will be able to use the model that suits it, by quality, cost or where its data may go. Sensitive work will be able to stay on models that run privately. If a provider changes its terms, the work will be able to move to another.
What this means for you
- You will not be tied to one AI vendor.
- Each task will be able to use the model that suits it best.
- Sensitive data will be able to stay on models that run privately.
Task: summarise client contracts · chosen because the data stays private
Your own deployment
Each client will get its own isolated deployment of Helm, never shared with another client. This is a fixed design decision.
- Your business
Its own deployment
HelmIts dataIts memoryIts gates - Another business
Its own deployment
HelmIts dataIts memoryIts gates - Another business
Its own deployment
HelmIts dataIts memoryIts gates
How it will be offered
Helm is not available yet. When it is, it will be offered in two ways.
A full deployment
Helm set up for your business alone, with its agents, gates and memory working on your projects. It would run in its own isolated deployment, connected to the tools you choose.
Rented Helm Agents and workflows
Individual Helm Agents or workflows for one specific job, without a full deployment. You would choose the job, and a person would still approve what matters.
Interested in Helm?
Helm is not available yet. If it sounds like something your business would use, register your interest and tell us what you would want it to do.