Frequently asked questions

Fabriq is a workspace OS shell for teams that ship code with AI agents. It sits between your repositories and your AI agent (OpenCode or Claude Code), intercepting every capability call and surfacing it for human review. It's not an IDE — it's a control panel for AI-assisted development.

Download the latest release from the download page. Fabriq runs on macOS, Windows, and Linux. No configuration is required — just point it at your workspace directory and attach your agent.

Yes. Fabriq is open source under the MIT license. You can use it, modify it, and distribute it freely — in source or binary form.

Fabriq uses the ACP (Agent Communication Protocol) stdio JSON-RPC bridge. It works with OpenCode, Claude Code, and any ACP-compatible agent. No PTY shims are needed.

No. Fabriq layers on top of your existing git workflow. You still use your preferred agent and your existing repository structure. Fabriq adds visibility and control without replacing your toolchain.

Fabriq enforces a two-dimensional workspace: long-lived workspace assets (specs, contracts, tokens) live on one dimension; short-lived feature worktrees live on another. The shell enforces this boundary at the routing level, the IPC layer, and in the UI. The agent can never accidentally cross-write.

Yes. Every state-changing capability call surfaces as an explicit confirmation prompt in the UI. You see the target dimension, the payload, and the expected effect. Nothing hits disk until you approve. Reversal is a first-class operation.

Every agent decision leaves sourced, validated evidence. Cognitive artifacts carry confidence levels (fact / inference / unknown) with mandatory evidence fields. The audit log is JSONL format with timestamps, capability names, durations, and source references. You can trace every change back to its origin.

Absolutely. Solo developers benefit the most — without a team to review your code, Fabriq gives you the confidence that your AI agent isn't doing something unexpected. See what your agent wants to change before it happens.

No. Confirmation prompts take seconds to review, and they're only shown for destructive operations. Reading diffs is optional, not mandatory. Fabriq is designed to add safety without friction — you can approve or deny capability calls with a single click.

Yes. Reversal is a first-class operation in Fabriq — not a post-mortem recovery. If an agent action causes issues, you can revert it through the UI.

The Feature Workbench provides a stage-gated lifecycle: from spec to solution design to implementation to validation to close. Each stage produces structured artifacts with evidence. You can see what decisions were made and whether the implementation matches the spec.

Yes. Every stage produces artifacts with provenance. The CDR (Cognitive Discovery & Reverse-engineering) pipeline extracts business rules, state machines, and behavior flows from the code. These artifacts carry evidence chains showing exactly where each claim came from.

Every state change goes through a typed, validated capability channel with a strict allowlist. If a capability isn't in the allowlist, it doesn't execute — even if the agent is compromised. The renderer never directly calls the engine; all communication goes through typed IPC with schema validation.

Fabriq itself is not SOC 2 certified (it's an open source tool, not a SaaS platform). However, the audit trail, evidence engine, and capability controls it provides are designed to help your team meet compliance requirements when using AI agents.

Fabriq is a local desktop application. All data — workspace assets, feature worktrees, audit logs, cognitive artifacts — is stored on your local machine. We don't have cloud servers, telemetry, or data collection. The only external communication is GitHub API calls for release data on the download page.

Electron 35 + React 19 + TypeScript for the desktop shell. Tailwind CSS 4 for styling. The engine (dapei-engine) runs as a subprocess or inline. Agents connect via ACP stdio JSON-RPC. The website is Next.js 15.

Yes. Fabriq has a plugin manifest system where routes, sidebars, feature panels, and agent backends register through a validated manifest at load time. The shell knows what's loaded before anything executes.

Fabriq is designed as a desktop GUI application, not a CLI tool. However, the deterministic validation engine can run in CI environments independently. Contact the team on GitHub if you have specific CI integration requirements.

Fabriq is not an IDE or an editor. It's a workspace OS shell that orchestrates external AI agents. Unlike IDE extensions that run inside the editor, Fabriq sits at the workspace level — managing multiple repositories, enforcing dimension boundaries, and providing a confirmation-driven UI for agent actions.