| Activity | The block above an answer showing what the agent did on that turn |
| Agent | The assistant that talks. It has instructions, provider, model, files and tools. See Agents |
| API key | The key an external system uses to call an integration or a folder |
| Cards | Components that display formatted content in the chat |
| Credential | A secret stored securely and reused by tools and MCP servers |
| Folder | Groups solutions, with shared memory and API key |
| Inherited | Marks a memory that comes from the folder or the solution, not from the agent itself |
| Instructions | The text defining the agent’s behaviour: language, tone, scope, what to refuse |
| Integration | The operation mode where the solution serves another system via webhook, with no chat page |
| MCP server | External server that publishes a catalogue of tools for the agents |
| Memory | What the agent keeps between conversations |
| Operation mode | WebChat or Integration |
| Orchestration | How MarIA picks which agent answers each message |
| Provider | The supplier of the AI model the agent uses |
| Access role | A set of permissions assigned to people and linked to solutions |
| Scope | A memory’s visibility: shared, individual or per conversation |
| Slug | The solution identifier used in the address; it never changes after creation |
| Solution | The workspace for one business subject |
| Specialist | An agent’s sub-assistant, consulted when a question needs depth |
| Tool | A capability beyond talking: call an API, show a card, collect a form |
| Trace / trace narration | The plain-language account of what happened on a turn |
| Vectorisation | Indexing a file for semantic search |
