ClearSpecs AI, offline: AI-assisted work items that never leave your environment

Most teams that want AI inside Azure DevOps run into the same wall before they ever write a prompt: the data isn't allowed to leave the building.

That's not a hypothetical for me. At FJAN IT I build tooling for teams whose work items contain things they cannot ship to a third-party SaaS: defect details, security findings, contract scope, patient or citizen data described in plain text. For those teams, "just connect it to a cloud AI service" is a non-starter, no matter how good the feature is.

That's exactly why ClearSpecs AI ships an offline, self-hosted version. Same extension, same AI-assisted work item authoring, but it runs entirely inside your own environment and talks only to an AI endpoint you control.

This post covers what the offline version is, who it's for, and how it actually works so the data stays put.

Why an offline version of ClearSpecs AI exists

The standard ClearSpecs AI extension uses a managed AI backend, which is the right choice for plenty of teams. But for many serious Azure DevOps users, the blocker has nothing to do with where Azure DevOps runs; it's about where the AI calls go. A managed backend is simply off the table when:

• Regulated data can't go to an external AI service. Healthcare, finance, defense, and public-sector teams often operate under rules that forbid sending work-item content to any third-party service.
• Data residency and sovereignty matter. You may be required to keep everything inside a specific tenant, region, or sovereign cloud, including the AI calls.
• Restricted or air-gapped networks rule out reaching any backend you don't operate yourself.

The offline version solves this at the boundary that actually matters: nothing in your work items is sent to FJAN IT or any FJAN-operated service. The AI calls go to an endpoint you host and own. And this is independent of where your Azure DevOps lives: it works the same on hosted Azure DevOps Services and on Azure DevOps Server on-premises.

Who the offline version is for

You're a strong fit for self-hosted ClearSpecs AI if you recognize yourself here:

• Government and public sector teams that need AI assistance without any citizen or case data leaving the network.
• Defense, aerospace, and critical infrastructure teams operating in restricted or air-gapped environments.
• Healthcare and finance teams bound by data-handling and residency rules.
• Anyone who can't route work-item content through an external AI service, whether you're on hosted Azure DevOps Services or Azure DevOps Server on-premises. The offline version supports both.
• Sovereign-cloud and single-tenant setups where everything has to stay inside your own Azure subscription.

If your blocker has always been "we love the idea, but compliance won't let the data leave," this is the version that removes the blocker.

How the offline version works (bring your own AI endpoint)

The design principle is simple: ClearSpecs AI should never be the thing that moves your data. You bring the AI; the extension just orchestrates the prompts.

In practice, you point ClearSpecs AI at your own Azure OpenAI or Azure AI Foundry deployment. The extension speaks the standard OpenAI-compatible API, so configuration comes down to three things you already control:

• Your endpoint: your Azure AI Foundry resource, exposed via its OpenAI-compatible path.
• Your deployment name: used as the model parameter, so you decide which model is in play.
• Your key and access policy: the credentials stay in your environment.

What stays inside your environment

• The work items being drafted, refined, or summarized.
• The prompts and any custom prompt templates you've defined.
• The model responses.

All of it travels between your Azure DevOps and your AI endpoint. FJAN IT is not in the path.

A practical note on browser-to-endpoint calls

Because the extension runs in the browser, calls to your AI endpoint cross a CORS boundary, and Azure AI Foundry doesn't expose a configurable CORS surface of its own. The clean, production-ready pattern here is to front your endpoint with a thin proxy, either Azure API Management or a small Azure Function, that handles CORS and keeps your key off the client. It's a one-time setup, and it keeps the whole flow inside your tenant. If you're standing this up, I'm happy to share the exact configuration that works.

You don't lose any features

The offline version isn't a stripped-down edition. It's the full ClearSpecs AI feature set, running against your own model:

• Draft and refine work items with AI, guided by your own custom prompts.
• Per-project prompt templates, so each team works the way it's supposed to, included in the license, not bolted on as an add-on.
• Work item summarization for long, sprawling tickets.
• Ad-hoc prompts for one-off "rewrite this as acceptance criteria" style requests.

Same workflow your team would get from the marketplace version, just with the data boundary you actually need.

How to get the offline version

The self-hosted option isn't a checkbox in the marketplace listing; it's set up with you so it fits your environment, your endpoint, and your compliance requirements. That's intentional: these deployments are rarely identical.

If the offline version sounds like the unlock you've been waiting for, schedule a session and tell me about your setup: whether you're on hosted Azure DevOps Services or Azure DevOps Server, and what your AI endpoint looks like. I'll walk you through exactly how ClearSpecs AI runs against your own endpoint, with your work-item data staying entirely on your side.

ClearSpecs AI: the same AI-assisted work items, on your terms, inside your walls.