Core concepts
This page is a quick map of the Imaginne vocabulary. Each term points to a full explanation.
The basics
| Concept | What it is |
|---|---|
| Surface | The interface you use Imaginne through: Desktop, terminal (TUI), VS Code, or chat in the browser. |
| Engine | The agent itself — it reasons, picks tools, reads and writes files, and runs commands. See Architecture. |
| Gateway | The model broker: the only component that talks to the AI models. See Models. |
| Platform | The control plane: identity, policy, and governance. It never runs your tools. |
| Project / workspace | The folder the agent works in. Files, memory, and sessions live there. |
Conversation and context
| Concept | What it is |
|---|---|
| Session | A continuous conversation with the agent — history, actions, and state. It's saved locally and can be resumed. See Sessions. |
| Turn | One round: your message → the model reasons → tools run → response. |
| Memory | Four Markdown notes per project (profile, project, glossary, decisions) that the agent reads at the start of the session. See Memory. |
| Artifact | A file generated by the agent — .pptx, .docx, .xlsx, .pdf, an image, code. See Files & artifacts. |
Capabilities
| Concept | What it is |
|---|---|
| Skill | A capability package the agent uses when relevant. See Skills. |
| Bundled skill | The document skills that are always available: docx, excel, pdf, ppt_skill. See Bundled skills. |
| Skill execution mode | local normal (open files) or local protected (verified bundle). See Execution modes. |
| Model | The AI that responds. Imaginne exposes nnumbers (primary, for general work) and nnumbers-code (a specialist for development and code). See Models. |
Control and security
| Concept | What it is |
|---|---|
| Autonomy | How much the agent asks before acting: low, medium (default), or high. See Autonomy. |
| Identity | You log in with your organization account and receive a single authenticated session, valid across every surface. See Identity. |
| Organization & policy | Your organization is the tenant; policy governs models, skills, credentials, and more. See Organizations & policies. |
| Profile | A set of skills and rules assigned to users. Without a profile, the chat opens empty. See Profiles. |
| Skill group | A way to organize skills and delegate their administration. See Skill groups. |
| BYOK | "Bring Your Own Key": the organization uses its own provider key. See BYOK. |
| Env-secret | An organization secret injected only into the skills that declare it. See Env-secrets. |
Ready to get started
With the vocabulary in hand, head to the Quickstart.
Was this page helpful?
Report a problem on this pageDo not send passwords, keys, tokens, or customer data.