Skip to main content
This guide walks you through setting up MIRA for the first time and running your first query with a concrete example at each step.

​
Before you start

Make sure MIRA is installed and the status bar shows NAE ready. If it shows Initializing, wait for the one-time setup to complete (~60 s). If it shows Failed, see Engine Settings to diagnose.

​
Step 1 — Connect a provider

Open Settings (⌘, on macOS / Ctrl+, on Windows/Linux).
Click the Bedrock / AWS tab and enter your credentials:
FieldExample
Access Key IDAKIA...
Secret Access KeywJalr... (shown as ●●●●)
Regioneu-west-1
Click Test Connection. On success, MIRA displays your AWS Account ID and IAM ARN — confirming credentials are valid and the Bedrock API is reachable.
Credentials are stored in your OS keychain (macOS Keychain / Windows Credential Manager / libsecret). They are never written to disk or log files.
Click Save and close Settings.

​
Step 2 — Choose your engine

The engine selector is in Settings → Engine. Two cards are shown:
  • Native Agent Engine (default) — multi-agent orchestration, adaptive memory. Best for open-ended questions, long sessions, and multi-domain reasoning.
  • RLM Engine — code-executing recursive reasoner. Best for data analysis, calculations, and tasks where a verified answer matters most.
For your first session, leave it on Native Agent Engine.

​
Step 3 — Start a chat

  1. Click New Chat (⌘N) in the sidebar
  2. An empty chat window opens
  3. Type a question. For example:
Summarise the key differences between transformer and state space models,
focusing on their computational complexity and what kinds of tasks each
architecture is best suited for.
  1. Press Enter (or click the send button)
Watch MIRA think in real time. You’ll see:
  • A streaming response appearing token by token
  • A status update in the Activity Center (⌘4) showing the query is in flight
  • The session title auto-generated from your question

​
Step 4 — Upload a document (optional)

Click the Paperclip icon in the message input toolbar (or press ⌘D) to open the Document Panel on the right side of the chat window.
  1. Drag and drop a PDF, DOCX, CSV, or text file onto the panel — or click Upload
  2. Wait for the status to change from Parsing to Ready
  3. The document is now automatically included in your next query
Ask a follow-up question about the document’s contents: 
Based on the uploaded document, what are the three most important points?
MIRA injects the document content into the reasoning context automatically.

​
Step 5 — Explore the views

Use the icon bar on the left to navigate:
ViewShortcutWhat’s there
Chat⌘1Your conversation
Skills⌘2Reasoning personas — try applying “Financial Analyst” to a new session
Workflows⌘3Multi-step pipelines — try “Research Report Pipeline”
Activity⌘4Real-time in-flight query monitor
Evals⌘5Evaluation framework for systematic quality tracking

​
What’s next?

Apply a Skill

Five built-in domain experts ready to use

Run the Research Pipeline

Multi-step research → analysis → report

Add an MCP Tool

Connect web search, databases, or custom APIs

Open the REPL Console

See every reasoning step in real time