Skip to main content

Screenshot gallery

A curated tour of the AIFactory portal, captured from the live deployment at aifactory.freundcloud.org.uk. These are real projects and real tasks — both a successful run and one that ended in errors — not mock-ups.

Refresh these with npm -w apps/frontend-web run capture-screenshots (drives the live portal via Playwright; see scripts/capture-screenshots.ts).

Getting in​

The portal authenticates with an API token, or with single sign-on via your organisation's identity provider.

API-token login screen, with the SSO option below it

The board​

The kanban board lays a project's tasks across the four PARR lanes — Plan (PFactory), Code (AIFactory), Review (TFactory) and Done (shipped). Each card carries its phase, a progress bar, and the Plan / Code / Validate / Log controls.

Kanban board showing tasks across the Plan, Code, Review and Done lanes

Creating a task is a single dialog: describe what you want in natural language (you can @-reference files), optionally name it, and the planner takes it from there.

Create New Task dialog

A task, in detail​

Opening a card gives you the full task: the spec on the Overview tab, the planner's Subtasks, streamed Logs, the changed Files, and an Observability tab.

Task detail — Overview tab with the generated spec

The Subtasks tab is the implementation plan — file references and per-step status, reviewable and editable before any code is written.

Task detail — Subtasks / plan tab

Logs stream the agent's work as it happens.

Task detail — Logs tab

Files shows what the run actually changed.

Task detail — Files tab

Observability breaks token usage down by category — system prompt, user messages, team/coordination context, tool outputs, thinking + assistant output — each with its own cost, plus resource usage and a direct "send to agent" channel.

Task detail — Observability tab with per-category token usage and cost

When a run goes wrong​

Not every run succeeds, and the board doesn't hide it. Tasks that hit problems carry a Has Errors badge and a failed progress bar.

A project board showing tasks in a Has Errors / failed state

Open one and you get the same detail surface — here a task sitting at the Human Review gate with errors flagged, the plan-review panel, and a Request Changes box to hand corrections back to the agent.

Detail of a task flagged with errors, at the Human Review gate

Beyond the board​

The portal is more than a kanban. The left rail carries the workspace tools.

The Files browser exposes the task's workspace.

Workspace Files browser

Agent Terminals spawn multiple terminals to run agents in parallel.

Agent Terminals view

MCP lists the Model Context Protocol servers available to agents.

MCP servers view

Worktrees tracks the git worktrees the build uses.

Worktrees view

GitHub Issues syncs issues straight into the board.

GitHub Issues sync view

Settings and providers​

Settings carries appearance, the default agent profile and model, integrations, git credentials, scoped API keys, and more.

The Agent Settings pane sets the default model and agent framework, with presets for complex tasks, balanced work, and quick edits.

Settings — Agent Settings, default model and agent profile

LLM Providers is where multi-provider lives: add multiple Claude subscriptions to auto-rotate on rate limits, connect Codex (OpenAI) and Antigravity (Google) CLIs, and point at OpenAI-compatible endpoints.

Settings — LLM Providers, cloud agents and OpenAI-compatible endpoints

Integrations holds third-party service connections and OAuth credentials.

Settings — Integrations and OAuth credentials