Halofy and its connected apps#
| Term | Meaning |
|---|---|
| Halofy | One platform with two connected apps. Inside the desktop app, the AI Workstation is simply called Halofy. See What is Halofy? |
| AI Workstation | The desktop app for macOS, Windows and Linux where you work with AI agents. See AI Workstation overview. |
| Console | The web app at app.halofy.ai where a person or an organization manages what agents know and may do, and reviews their work. See What is the Console? |
| Halofy account | The account you sign in to under Settings → Linked devices → Halofy account. It links your devices and connects the app to your Console workspace. The Console has its own sign-in at app.halofy.ai; Halofy matches the two by verified email address. See Sign in to your Halofy account. |
| Workspace | Your space in the Console. It's a personal workspace (For myself) or an organization workspace (With my organization or team), chosen when you create it. |
| Team | A group of people in an organization workspace. Skills, policies, instructions and allowed MCPs can be set for each team, and members use the memory granted to their teams. See Teams. |
| Scope | The part of a workspace a setting applies to: the whole organization or one team. A team also receives what's published for the organization above it. |
In the AI Workstation#
| Term | Meaning |
|---|---|
| Provider | An AI service whose agent Halofy runs, such as Claude, Codex, Cursor, Antigravity, Grok, Grok Bot, Pi, OpenCode, Kimi Code, Muse, DeepSeek or Hermes. Connect providers in Settings → AI services. See Connect your AI providers. |
| Provider account | One named connection to a provider, added with Add provider in Settings → AI services. You can add more than one account for the same provider. |
| Project | A folder, often a code repository, that your agents work in. See Projects and project settings. |
| Thread | One conversation with an agent inside a project, together with its work history. Stopping the agent doesn't delete the thread. See Threads and the sidebar. |
| Turn | One cycle of you sending a message and the agent working on it. |
| Worktree | A separate Git checkout that a thread can use, so its changes stay on their own branch. Choose New worktree when you start the thread. |
| Checkpoint | A saved state of the project's files at a point in a thread. Halofy uses checkpoints to show what changed, and to restore the files when you edit and resend an earlier message. |
| Permission mode | How much an agent may do without asking: Supervised, Auto-accept edits, Auto or Full access. See Permission modes. |
| Work view, Code view | Two ways to show the app. Work hides terminal, Git and diff tools; Code shows everything. See Work and Code views. |
| Environment | A computer running Halofy that your app is connected to, such as this computer or a remote machine. Listed under Settings → Linked devices. |
| Split view | Several threads shown side by side in panes. See Split view. |
| Settled, Snoozed | Thread states in the sidebar. Settling moves a thread to the Settled shelf. Snoozing moves it to Snoozed until the time you pick, then it wakes. |
| Tool activity | The actions an agent takes while it works, such as reading files or running commands, shown in the conversation. See Follow thinking and tool activity. |
| Skill | A reusable set of instructions an agent loads for a particular kind of task. Type $ in the message box to add one. Team skills come from the Console. |
| Shared instructions | Standing instructions under Settings → AI services that Halofy gives to Claude and Codex. Other providers get them only if you add them to a project's AGENTS.md. See Shared instructions and agent context. |
| AGENTS.md | A file in a project with instructions for the agents that work there. Most providers read it; Claude reads it when the project's CLAUDE.md imports it. |
| Artifact | A file an agent delivers into the conversation for you to preview or download. See Artifacts. |
| Fusion model | A saved team of models from your connected providers: a coordinator plus specialists with their own rules. See Fusion Reactor. |
| SnapShot | A capture of the window you're working in, attached to your draft. See SnapShots. |
People, roles and sharing#
| Term | Meaning |
|---|---|
| Owner | The one person with full control of an organization workspace: teams, membership, roles, access and organization settings. Whoever creates the workspace becomes its owner. See Roles and permissions. |
| Admin | Manages teams, people and invitations across the organization, but can't change the owner. |
| Member | Works in their assigned teams and uses the memory granted to those teams. |
| Invitation | An email invite or link, sent from Teams, that lets a person join a team. Invite links expire after 7 days. |
| Team activity | The sharing status for new conversations and supported activity from a connected team computer. Connecting to the team enables sharing; signing out stops it. See what connecting shares. |
| Agent badge | The credential an agent uses to reach the Console, scoped to one team. Managed under Scope Access. |
In the Console#
| Term | Meaning |
|---|---|
| Setup | The Set up Halofy checklist a new workspace completes: add skills, add policies, choose allowed MCPs and add agent instructions. See Sign in and set up your workspace. |
| Overview | The Console's home page. It shows captured agent activity at a glance and lets you analyze sessions for practices worth reusing. See The Overview page. |
| Skills | The library of reusable methods available to your teams. Uploaded skills can be available while background checks report advisory findings; generated playbook drafts still follow publication review. See current skill checks. |
| Policies | Policy files, such as text, Markdown, CSV, JSON or YAML, that apply to a scope and its child scopes and are delivered to authorized agents. See Policies. |
| Instructions | Versioned operating instructions for supported agents, published per organization or team and delivered when agents sync. See Instructions. |
| MCPs | The Console page where you choose which MCP servers agents may use, or allow none. See MCPs. |
| Scope Access | Where you define scoped access for a team and manage agent badges. See Scope access. |
| Memory | What your workspace has saved, organized by team. See Memory. |
| Agent Connections | Where you connect agents to the Console and see each connection. Workstations appear as Halofy AI Workstation. See Agent connections. |
| Assistant | A chat in the Console that answers questions from your workspace, using a model set up in AI models. See Assistant. |
| Sessions | Captured agent conversations. Labelled Agent sessions in a personal workspace and User sessions for organization managers. See Sessions. |
| Audit | A log of every read, write, deletion and export of your organization's memory. See Audit log. |
| Export | A download of captured sessions' details and summaries, for owners and admins. See Export. |
| AI models | Where you connect model providers and deploy the models the Assistant uses. See AI models. |
Data and records#
| Term | Meaning |
|---|---|
| Captured session | An agent conversation shared with the Console, through team activity or a connected agent. |
| Session summary | A short summary the Console creates for a captured session. |
| Retention period | How long captured conversations are kept, shown as Keep conversations. The default is 365 days; an owner can choose 30, 90, 180, 365 or 730 days, or Indefinite. |
| Legal hold | Stops a captured session from being deleted when its retention period ends. Releasing the hold applies the retention period again. |
| Erase | Permanently removes a captured session and what was derived from it. It can't be undone. |
| Estimated spend | An API-equivalent cost estimate calculated from reported token usage and a price catalog. It isn't your provider's bill. |
| Product usage data | Events the Workstation sends about how the app is used, without your prompts or files. See Your data and privacy. |
Playbooks, setup and usage in plain language#
- Playbook: a reusable method with inputs, steps and result checks. Published playbooks reach agents through the skills library. Playbook guide.
- Guarded run: an optional separate conversation prepared with a playbook’s approved setup and tool restrictions. It does not guarantee every written instruction is followed.
- Reference-only: a playbook you can read and follow as guidance, without starting a guarded run.
- Partial setup: some source files or settings were not collected. The available setup can still run when its other checks pass; review the omissions.
- Current setup: supported instructions, skill text and safe settings collected from a source computer now. It is separate from captured chats and does not establish what an old chat used.
- Power user workflow: an inferred label on a chat showing a reusable way of working. It is not a rating of the person.
- Token: a small piece of text a model processes. Token-based cost estimates are not your subscription bill.
- CSV: a spreadsheet-friendly download. The Console’s chat export has one row per captured chat, with summaries and available usage.
Personal workspace owners can use Audit and Export. Uploaded skills have advisory background checks, and members can view their assigned Policies, Instructions and MCPs. See the account guide and skill-check guidance for details.
