Mira User Manual

Back to Mira
MiraMira

03 Core concepts and features

Projects · Conversations · Tasks

Projects at a glance

A project is the basic unit of work in Mira. Each project has its own knowledge base, task list, and conversation space for a research topic or an ongoing line of work.

These three elements form a connected structure:

  • Projects define the scope of a particular research effort.
  • Conversations are the direct entry point for everyday collaboration with Mira.
  • Tasks handle complex work through structured execution initiated from a conversation.

When a task finishes, its key findings and outputs can be added to the project knowledge base for reuse. Work accumulates around the project, building context and results that you can draw on and extend in later tasks.

This structure keeps different research topics, source files, and stages of work clearly separated. For research that continues over time, with new materials and intermediate results, a project brings related work into one consistent workspace.

Create a project

Your current project name appears below the Mira logo at the top left. Click it to open the project list or create a project. Keeping different research topics in separate projects gives each project a clear memory boundary and helps the agent understand its context.

Creating a project involves the following steps:

  1. Choose whether the project runs locally or in the cloud. This choice cannot be changed later. A local project lets Mira read local files, execute commands, and work in your computer's environment. Conversations in a local project are saved on that computer and do not sync to other devices.

Mira Web supports remote projects and cannot create local projects.

  1. Enter a project name and description. A specific description helps Mira work toward your goals.
  1. For a local project, select a working directory. Files created in its conversations are saved there.
  2. Optionally select Reference other projects to reuse their knowledge bases. You can add references after creating the project too, so results collected in one project can inform work in another.
  3. To edit an existing project, open the project list and click the “…” button beside its name. You can change its name, description, and references to other projects.

Conversations

Conversations are the main way to work with Mira.

Almost every capability starts in a conversation: asking a question, analyzing a file, calling a skill, starting a complex task, or developing research findings.

Think of the conversation as the shared entry point to Mira's workflows. Mira answers simple requests directly. When a request needs a more complex process, it starts a task from the conversation and continues the work there.

You can create as many conversations as you need within a project. Each conversation has its own context window and does not automatically inherit the history of other conversations. This lets you work on separate questions in parallel without mixing discussions.

To use materials already collected in the project, turn on the knowledge base switch to enable retrieval-augmented responses based on the project knowledge base.

Start a conversation

Create a new conversation when you want to explore a new question, idea, or task.

Click New conversation near the top of the sidebar to begin an independent conversation.

A new conversation starts with a fresh context window and is useful for:

  • Starting a new research discussion.
  • Working on different tasks in separate conversations.
  • Keeping earlier discussion from influencing the current question.
  • Exploring separate subquestions within the same project.

In a connected messaging channel, send /new to the bot to start a new conversation.

View and delete conversation history

Previous conversations remain in the history list. Open the conversation title menu to search, switch between, or delete them.

A deleted conversation is removed from the current project's conversation list. Use this to organize your workspace, remove temporary tests, or clear discussions you no longer need. Deleting a conversation does not affect other conversations, tasks, or knowledge base content in the project.

Use skills in a conversation

Many Mira capabilities are triggered by instructions in a conversation.

To use a specialized capability, type / in the input box, open the skill list, and choose a skill.

For example, ask Mira: “Use /deep-research to study tools and models related to AI scientists and analyze future trends.” Mira will use the deep-research skill for the task.

Skills are callable capability modules for more specific, specialized, or structured work, such as analysis, file processing, data operations, or working with external tools.

You can use skills in any of these ways:

  • Type “/” in the input box and select a skill from the list.
  • Describe your goal in the prompt and let Mira decide whether to call a suitable skill.
  • Click Skills below the input box and select the skill you need.

This keeps your work centered on the conversation and reduces the need to move between menus.

Add files to a conversation

Upload files directly into a conversation and discuss them with Mira.

Mira accepts most common file formats, including documents, spreadsheets, presentations, source code, datasets, images, structured text, and other research materials.

Use uploaded files to:

  • Ask Mira to read and explain their contents.
  • Extract key points, summarize structure, or answer questions.
  • Continue with analysis, comparison, or organization.
  • Provide inputs for a later task.

Files uploaded to a conversation are attachments to that conversation. To keep a file as a reusable project resource, upload it through the knowledge base or use Add to knowledge base where available.

Cloud conversations have a per-file upload limit: 500 MB for Ultra and 100 MB for Free through Max. Local projects have no file upload size limit.


Tasks

When a request needs complex tool use, multistep analysis, or a longer execution process, Mira creates a task within the conversation.

A task is a distinct execution process for work such as combining sources, multistep analysis, extended reasoning, complex tool calls, or producing a complete result.

Tasks have several key properties:

  • They start from a conversation and are recorded as independent units of work within it.
  • They produce structured findings that can be reused and traced later.
  • They can be stopped at any time; output files already created remain accessible.
Task management

Conversations and tasks appear under Project conversations in the sidebar, organized as Conversation → Task → Output file. Expand a conversation to see its tasks, then expand a task to see its files. Search outputs at the top of the tab to search names across conversations, tasks, and files.

You can also browse the current conversation's tasks and output files in the outputs card at the upper left of the conversation area.

Hover over a task and open its ⋯ menu to:

  • Go to source: Return to the message that started the task.

For cloud tasks, use Delete folder to remove the task's folder and outputs.

Work with output files

Hover over an output file for quick actions:

  • Click the filename to preview it in the right panel.
  • Reference in conversation: Click the @ icon or drag the file into the input box to attach it as a reference.

Add cloud project files to the knowledge base for future use. A green checkmark indicates a file that is already in the knowledge base.

Manage downloads, deletion, and indexing status in the Knowledge base tab, where files become searchable throughout the project. See the Knowledge base section.

Task details and outputs

When a task finishes, its execution process, report, and key files usually appear in the main conversation. Click View task details and outputs to expand a task card containing:

  • The task name in the current conversation.
  • Task status.
  • Run time.
  • Task outputs.
  • Number of outputs.
  • A button to add outputs to the knowledge base in bulk.

The execution process helps you understand the model's reasoning, the tools it used, the files it produced, and how it reached the result.

MiraMira