PRANUS

DEVUS

Custom features for your business, without the developer backlog

DEVUS is the AI development team inside PRANUS. You describe what’s missing, and a team of AI roles takes it from task to release — spec, tests, code, checks — and ships only when you approve.

What it does

  • Development requests start as tasks

    Every task in TASKUS has a Development tab. Describe what you need and follow progress step by step, without leaving the task.

  • A team of roles, not a single bot

    A scout studies the system, an analyst writes the spec, a critic pokes holes in it, a builder writes the code, and an independent reviewer checks it with fresh eyes.

  • Tests first, then code

    Failing tests are written before any code. The work is done only when they pass, and every requirement is covered by a test.

  • No faked green tests

    Tests are locked before work begins, so the code can’t quietly bend them. Passing tests on a broken core feature never count as done.

  • Checked in a real browser

    Before handoff, DEVUS opens the result like a user would and confirms the main action actually works — not just that the build succeeded.

  • A sandbox for every job

    Each piece of work runs in its own isolated environment. Your live system doesn’t see the change until it’s accepted.

  • Ships only with your approval

    A human approves the spec and the release. After release, the change is verified and rolled back automatically if something fails.

  • The lightest fix first

    DEVUS climbs a ladder: configuration, then data definitions, then workflows, and only then sandboxed code. The lower the rung, the simpler and safer the change.

“We just need one more field” — and a six-month wait

Every business system eventually runs into the way your company actually works. You need a report the template doesn’t have. A field on the customer record that the warehouse can’t ship without. A check your accountant currently does by hand in a spreadsheet. You email your vendor, get a quote, and wait in line. Three months later the change arrives — and breaks something that used to work.

Most mid-sized companies don’t have an in-house engineering team, and those that do have a backlog booked a quarter ahead. So small improvements never happen, and people learn to work around the system with spreadsheets and chats.

DEVUS removes that line. You describe the task in plain language, and an AI development team handles it the way a careful human team would: understand first, then write a spec, then tests, then code — and ship only with your approval.

How DEVUS connects to the rest of PRANUS

  • Tasks. Work for DEVUS is a regular task in TASKUS task management. Its card gets a Development tab with steps, approvals, and results.
  • ERP rollout. When the implementation agent configuring OPTIMUS ERP hits a capability that doesn’t exist yet, it can file a request with DEVUS instead of saying “that’s not possible.”
  • Websites. WEBUS runs on DEVUS. Simple site edits happen in plain language, and developer mode opens the full DEVUS team on the same project.
  • Business rules. If a change involves a rule like “ship only after the delivery act is signed,” you can verify it in LOGUS before it goes into code.
  • AI teammates. Plenty of needs don’t require code at all — a new AI agent is described in plain words in the AGENTUS agent builder. DEVUS steps in when a description isn’t enough.

What the AI does

Every task follows the same transparent path:

  1. Risk triage. Is this a small tweak, or does it touch money and permissions?
  2. Scouting. The AI studies how the relevant part of the system works and what depends on it.
  3. Spec and critique. A spec describes exactly what will change, and a separate critic role looks for gaps and contradictions.
  4. Human approval. You read the spec in plain language and approve or adjust it.
  5. Tests first. A test is written for each requirement. With no code yet, they fail — as they should.
  6. Code and quality gates. The builder writes the code, and automated checks block progress until everything lines up.
  7. Browser check. The result is opened like a user would open it: does the main action work?
  8. Fresh review. A reviewer who didn’t write the code examines it independently.
  9. Release with consent. After your approval, the change ships, gets verified, and rolls itself back if something fails.

The rule that matters most: green tests on a broken core feature are not “done.” DEVUS proves it didn’t break anything instead of just saying so.

Getting started

DEVUS comes in after the core rollout, once your org chart, permissions, team, pipelines, and knowledge base are in place. When it turns out the standard setup isn’t enough, the task goes to DEVUS. See how PRANUS gets set up through a conversation with AI in implementation in 14 days.

Who it’s for

  • Companies without an engineering team that want the system to fit their processes, not the other way around.
  • IT teams that want to offload the stream of small requests and keep people focused on architecture.
  • Owners and managers tired of working around their software with spreadsheets.

Want to watch a task go from a plain-language request to a release? Book a demo and we’ll walk you through a real one.

FAQ

Do we need our own developer?

No. A manager or process owner describes the task in plain language. A person is needed only to approve the spec and the release, so the decisions stay with you.

Could the AI break something that already works?

Every job runs in its own sandbox, tests are written before code, and the result is checked in a browser before handoff. Nothing reaches your live system without your approval, and failed releases roll back automatically.

How is this different from AI app builders?

Most AI app builders generate a separate app that sits next to your core system. DEVUS changes the system itself — the one that already holds your people, customers, and money — and every change goes through the same checks as professional development.

What kinds of work can DEVUS take on?

Anything the standard setup doesn’t cover: a new field or section, a report, an automation, a connection to your own system, a screen for an unusual process. If it can be solved with configuration, DEVUS does that instead of writing code.

Who can see progress?

The task’s author and owners see each step on the Development tab: what’s done, what’s waiting for approval, and what’s being tested.

Let’s bring order in two weeks

Tell us about your company — we’ll show PRANUS on your real tasks and, if we’re a fit, implement it for free.