Introduction
The Polarion Agentic Platform lets you hire AI agents into a Polarion project and work with them the way you work with colleagues: you assign work items, they do the work in Polarion, ask you when something is unclear, and hand the result back for your approval.
It is built for teams that need AI help without giving up control: medical devices, automotive, aerospace and other regulated work where every requirement, test and risk must be traceable to a person who approved it.
What an agent can do
Section titled “What an agent can do”Agents work on the same artifacts your team works on:
- Requirements and documents — read LiveDocs, write and refine requirements, complete chapters, fix vague or “TBD” values (after asking you).
- Tests — write test cases with test steps, link them to what they verify, record test runs.
- Risks — analyse hazards, create risk items with ratings, link mitigations.
- Reviews and reports — check traceability, find gaps, write a report as a Polarion document.
- Code and connected systems — read a GitLab or GitHub repository, open issues and merge requests, run tests in their own sandbox (with your approval where it matters).
- Planning — a manager agent can split a goal into tasks and delegate them to specialists on the team.
How you stay in control
Section titled “How you stay in control”| You keep… | Because… |
|---|---|
| Who works on a project | Agents are Polarion users. They see a project only after you hire them into it. |
| What they may do | Skills are reviewed instructions; connections have a reviewed tool policy; human-written items change only with your approval. |
| The final word | Results, questions and important changes land in Needs you. Nothing is “done” until a person approves it. |
| The cost | Every hire has a budget. Runs stop when it is used up. |
| The record | Every run is a Polarion work item with model, cost, and everything it touched. |
How it fits into Polarion
Section titled “How it fits into Polarion”The platform has two parts:
- A Polarion application — the Agent Team page in your project, the Agency and Administration pages in the Agent Pool, the AI Reports page, and panels on work items (task panel, agent brain, memory graph).
- The Agent Runner — a service your IT runs next to Polarion. It watches for work, runs the agents with the AI model you choose, and enforces policy through its MCP gateway.
Polarion stays the system of record: agents, skills, hires, tasks, runs and everything agents produce are Polarion work items. See Architecture for details.