The launchpad for your AI operating model.
Quilane AgentOS is for launching and running your organisation's AI operating model: the agents, processes, skills, knowledge and rules that describe how each function works with AI. Start from a default OS, adapt it department by department, have your own owners approve every change, and publish one source to Claude and Copilot Studio.
Preview · limited access
Access is by request; Quilane opens each workspace.
Where AgentOS fits: steer, deploy, measure
Quilane covers two parts of the AI transformation. Quilane Cockpit, available from launch, is where leadership steers the impact and the spend. AgentOS is where your owners launch and run the operating model. People act at every step.
- Steer.
Quilane Cockpit
In Quilane Cockpit, leadership, through its AI council, decides what to build, scale, automate or stop. Frequent, proven, still manual use cases are flagged as agent candidates, and the council confirms or rejects each one with a reason.
- Build
- Scale
- Automate
- Stop
- Deploy.
Quilane AgentOSPreview
Your owners carry out the decisions to build, scale and automate in AgentOS.
- Measure.
Quilane Cockpit
The use cases those agents serve go on being measured in the Cockpit, on the same three questions: maturity, adoption and capture, with spend set against value.
The two products are separate. The loop is how your people use them, not a connection between them. AgentOS runs no agent, does not count the usage of the agents it delivers, and does not measure people's AI maturity.
Start from a default OS, on a shared foundation
Quilane supplies a default OS for an organisation's core functions. Activate it as is, or copy it and adapt it. Quilane updates are offered, never forced.
A global library holds what the whole organisation shares, and each department has its own library. Both hold processes, skills, knowledge, tools, rules, and the autonomy level of each process. When a second department uses a skill, the skill is promoted to the global library without changing its identity.
- Default OS: activate Quilane's default OS as is, or copy it and adapt it
- Existing agents: bring in your organisation's existing agents, through a migration kit
- A blank page: start from nothing and build your own
Global library
Processes, skills, knowledge, tools, rules, and the autonomy level of each process
Skill- Department librarySkill
- Department librarySkill
- Department library
Promoted to the global library when a second department uses it, without changing its identity
Your owners approve what goes live
Every change approved
Every change waits for approval by a department owner, or by the workspace administrator for the global library. Authors cannot approve their own change when another owner exists. Every approved version is kept; rolling back means republishing an earlier version.
A named owner for each agent
Each agent has a named owner and maintainer, chosen from your organisation's own directory.
Autonomy, declared in words
Each process declares its autonomy in words: draft, human approval, or bounded autonomy within a written scope. Moving to bounded autonomy is a manual, approved decision with a named owner. There is no fully autonomous level in the first release.
In Quilane Cockpit, maturity is measured. In AgentOS, autonomy is declared.
One source, published to your platforms
Publish everything, the global library, or one department. From one source, AgentOS publishes plugins for Claude and a managed solution for Copilot Studio. Both are placed in your organisation's own GitHub repository, in your test repository first if you have one. Each release carries a build identifier.
From there, your platforms take over. Claude syncs the plugins to your users. For Copilot Studio, your own pipeline installs the solution in test, then in production after your approval.
Quilane AgentOS
One sourceEverything, the global library, or one departmentEach release carries a build identifier
Your organisation, on your side
Your GitHub repositoryTest repository, if you have oneYour repository- Claude:plugins, synced to your users by Claude
- Copilot Studio:a managed solution, installed by your own pipeline in test, then in production after your approval
AgentOS runs no agent. Your agents run in Claude and Copilot Studio, on your side, and the people who use them never sign in to AgentOS. AgentOS holds no AI platform or Power Platform credentials.
Trust, requirements and the first release's scope
Each workspace isolated
Each organisation's workspace is isolated at database level. In the first release, Quilane has no access to the content of your workspace.
Every change traced, every write checked
Every change is traced: who, when, what, before and after. Deliverables are checked before every write: no AgentOS source, no secret, no personal name or identifier, no other customer's data. The Quilane GitHub app writes only to the repository your organisation authorised, with the minimum rights, and every write is logged. Without valid consent, nothing is written.
Export, and its limit
Your workspace administrator can export the full source at any time, with its history, in a readable format with integrity checksums. Deployed deliverables keep running. The export does not let you rebuild your deliverables without Quilane.
What you need, and how access works
Your organisation needs a Microsoft Entra ID directory and a GitHub organisation. People sign in with Microsoft Entra ID, and the workspace is derived from your directory. There is no self sign-up: access is by request, and Quilane opens each workspace.
Not in the first release
AgentOS publishes to Claude and Copilot Studio. The first release does not cover:
- Microsoft 365 Copilot declarative agents
- OpenAI
- Code hosts other than GitHub
- Automated answer-quality tests (evals)
Request access to Quilane AgentOS
Preview · limited access
Access is by request, and Quilane opens each workspace. Tell us about your organisation, and name the departments you want to start with in your message. We reply by email.