# MIRA

> MIRA — Multi-Step Intelligent Reasoning Agent. On-device Localized AI Agent.

## Docs

- [Changelog](https://docs.mira-app.dev/community/changelog.md): Full version history for MIRA.
- [Code of Conduct](https://docs.mira-app.dev/community/code-of-conduct.md): Our commitment to a welcoming and inclusive community.
- [Contributing](https://docs.mira-app.dev/community/contributing.md): Everything you need to go from zero to your first merged PR on MIRA.
- [FAQ](https://docs.mira-app.dev/community/faq.md): Answers to the most common questions about MIRA.
- [Roadmap](https://docs.mira-app.dev/community/roadmap.md): Planned features and improvements for future MIRA releases.
- [API Tokens & Security](https://docs.mira-app.dev/configuration/api-tokens.md): How MIRA stores credentials and what you can do to audit and rotate them.
- [NAE Settings](https://docs.mira-app.dev/configuration/nae-settings.md): Configure the Native Agent Engine — provider, model, context window, and sub-agent behaviour.
- [Preferences](https://docs.mira-app.dev/configuration/preferences.md): Personalise MIRA's appearance and behaviour.
- [Anthropic (Claude)](https://docs.mira-app.dev/configuration/providers/anthropic.md): Connect MIRA to Anthropic's Claude models.
- [AWS Bedrock](https://docs.mira-app.dev/configuration/providers/bedrock.md): Connect MIRA to Claude and other foundation models via AWS Bedrock — the default provider.
- [Ollama (Local Models)](https://docs.mira-app.dev/configuration/providers/ollama.md): Run open-source models locally with zero data egress using Ollama.
- [OpenAI](https://docs.mira-app.dev/configuration/providers/openai.md): Connect MIRA to OpenAI's API.
- [Reset & Data Management](https://docs.mira-app.dev/configuration/reset-data.md): Clear sessions, skills, workflows, eval data, or all MIRA data from your machine.
- [RLM Settings](https://docs.mira-app.dev/configuration/rlm-settings.md): Configure the Recursive Loop Machine — provider, model, iteration limits, and concurrency.
- [How MIRA Thinks](https://docs.mira-app.dev/core-concepts/how-mira-thinks.md): MIRA's approach to reasoning: why two engines exist and how they approach the same problem differently.
- [Native Agent Engine (NAE)](https://docs.mira-app.dev/core-concepts/native-agent-engine.md): How NAE orchestrates sub-agents, manages context, and maintains memory across long sessions — all with direct LLM API calls on your device.
- [Privacy & Data](https://docs.mira-app.dev/core-concepts/privacy-and-data.md): How MIRA keeps your data, documents, and credentials private — by design, not by policy.
- [Reasoning Engines](https://docs.mira-app.dev/core-concepts/reasoning-engines.md): Side-by-side comparison of MIRA's two engines — Native Agent Engine and RLM — with guidance on when to use each.
- [RLM Engine](https://docs.mira-app.dev/core-concepts/rlm-engine.md): How the Recursive LLM Engine writes and executes Python code, observes real output, and iterates until answers are verifiably correct.
- [Adding an Eval Type](https://docs.mira-app.dev/developer-guide/adding-eval-type.md): Extend the Eval Framework with a custom scoring strategy.
- [Adding a Document Parser](https://docs.mira-app.dev/developer-guide/adding-parser.md): Extend MIRA's document context with support for new file formats.
- [Adding a New LLM Provider](https://docs.mira-app.dev/developer-guide/adding-provider.md): Step-by-step guide to integrating a new AI provider into MIRA.
- [Architecture Overview](https://docs.mira-app.dev/developer-guide/architecture.md): High-level architecture of MIRA — how Electron, React, Python, and the reasoning engines fit together.
- [Build & Run (Development)](https://docs.mira-app.dev/developer-guide/build-dev.md): Set up your local development environment and run MIRA from source.
- [Build for Production](https://docs.mira-app.dev/developer-guide/build-production.md): Package MIRA as a signed, distributable desktop app for macOS, Windows, and Linux.
- [IPC Reference](https://docs.mira-app.dev/developer-guide/ipc-reference.md): Complete reference for all Electron IPC channels exposed through window.api.
- [Release Process](https://docs.mira-app.dev/developer-guide/release-process.md): How MIRA versions are cut, signed, and published — from first beta to stable GA.
- [Testing](https://docs.mira-app.dev/developer-guide/testing.md): Run the Python test suite and understand the testing strategy for MIRA.
- [A/B Comparison](https://docs.mira-app.dev/eval-framework/ab-comparison.md): Compare two eval runs side by side to measure the impact of a change.
- [Eval Dashboard](https://docs.mira-app.dev/eval-framework/dashboard.md): View conversation eval results, score trends, and A/B comparisons.
- [Rule Evals](https://docs.mira-app.dev/eval-framework/exact-match.md): Deterministic evals that check output against strings, regexes, keywords, schemas, and length constraints.
- [Exporting Eval Results](https://docs.mira-app.dev/eval-framework/exporting-results.md): Download eval results as CSV or JSON for reporting and archiving.
- [Human Review](https://docs.mira-app.dev/eval-framework/human-review.md): Manually override an automated eval result with a human score and note.
- [LLM Judge Evals](https://docs.mira-app.dev/eval-framework/llm-judge.md): Use a second LLM call to grade engine output against a custom judge prompt.
- [Eval Framework Overview](https://docs.mira-app.dev/eval-framework/overview.md): Test MIRA's reasoning quality systematically with built-in evaluation tools.
- [Eval Profiles](https://docs.mira-app.dev/eval-framework/profiles.md): Organise eval definitions into named, activatable groups.
- [Running Evals](https://docs.mira-app.dev/eval-framework/running-evals.md): How MIRA automatically evaluates every agent response against active eval profiles.
- [Semantic Similarity Evals](https://docs.mira-app.dev/eval-framework/semantic-similarity.md): Score engine output by measuring how close it is to a reference answer using TF-IDF cosine similarity.
- [Activity Center](https://docs.mira-app.dev/features/activity-center.md): Monitor in-progress queries and review a history of completed, failed, and cancelled requests across all sessions.
- [Session Management](https://docs.mira-app.dev/features/chat/session-management.md): Pin, rename, duplicate, and delete sessions. Understand how MIRA persists conversation history.
- [Starting a Session](https://docs.mira-app.dev/features/chat/starting-a-session.md): How to create a new chat session, optionally apply a Skill or Workflow, and begin reasoning.
- [Streaming & Markdown](https://docs.mira-app.dev/features/chat/streaming-and-markdown.md): How MIRA renders streaming responses — real-time tokens, rich Markdown, code syntax highlighting, and tables.
- [Document Chunking](https://docs.mira-app.dev/features/document-context/document-chunking.md): How MIRA splits large documents into context-sized pieces.
- [Supported File Formats](https://docs.mira-app.dev/features/document-context/supported-formats.md): Every file format MIRA can parse and inject into the reasoning context.
- [Uploading Documents](https://docs.mira-app.dev/features/document-context/uploading-documents.md): How to upload documents to MIRA and control which ones are active in each session.
- [Adding an SSE MCP Server](https://docs.mira-app.dev/features/mcp-tools/adding-sse-server.md): Connect a remote HTTP-based MCP server to MIRA using Server-Sent Events.
- [Adding a stdio MCP Server](https://docs.mira-app.dev/features/mcp-tools/adding-stdio-server.md): Connect a local command-line MCP server to MIRA.
- [Session vs Global MCP Servers](https://docs.mira-app.dev/features/mcp-tools/session-vs-global.md): Choose the right scope for each MCP server — active only in one chat, or always available.
- [What is MCP?](https://docs.mira-app.dev/features/mcp-tools/what-is-mcp.md): How the Model Context Protocol lets MIRA call external tools without writing integration code.
- [REPL Console](https://docs.mira-app.dev/features/repl-console.md): View the RLM engine's code execution output in real time.
- [Built-in Skills](https://docs.mira-app.dev/features/skills/built-in-skills.md): The five Skills that ship with MIRA — domain-expert reasoning personas ready to apply to any session.
- [Creating Custom Skills](https://docs.mira-app.dev/features/skills/creating-custom-skills.md): Build your own reasoning persona with custom instructions, model overrides, and MCP tool access.
- [Skill Import & Export](https://docs.mira-app.dev/features/skills/import-export.md): Share Skills as portable JSON files, restore previous versions, and collaborate across teams.
- [Skill Scope](https://docs.mira-app.dev/features/skills/skill-scope.md): Control whether a Skill applies to one session or automatically to every new session.
- [Building a Workflow](https://docs.mira-app.dev/features/workflows/building-a-workflow.md): Create a custom multi-step pipeline with MIRA's workflow builder.
- [Built-in Workflows](https://docs.mira-app.dev/features/workflows/built-in-workflows.md): Three production-ready multi-step pipelines that ship with every MIRA installation.
- [Workflow Import & Export](https://docs.mira-app.dev/features/workflows/import-export.md): Share workflows as portable JSON files and restore previous versions.
- [Routing Rules](https://docs.mira-app.dev/features/workflows/routing-rules.md): Control the flow of a workflow step — continue, retry, or skip based on conditions.
- [Step Types & Templates](https://docs.mira-app.dev/features/workflows/step-types-and-templates.md): Reference for the three step types and the template variables available in workflow steps.
- [Your First Session](https://docs.mira-app.dev/get-started/first-session.md): Walk through your first MIRA session — from connecting a provider to getting a reasoned answer.
- [Linux Installation](https://docs.mira-app.dev/get-started/installation/linux.md): Install MIRA on Ubuntu 20.04+, Fedora 36+, or any x64 Linux distribution.
- [macOS Installation](https://docs.mira-app.dev/get-started/installation/macos.md): Install MIRA on macOS 12 Monterey or later — Apple Silicon (arm64).
- [Windows Installation](https://docs.mira-app.dev/get-started/installation/windows.md): Install MIRA on Windows 10 or later — x64.
- [Introduction](https://docs.mira-app.dev/get-started/introduction.md): MIRA is a multi-step intelligent reasoning agent for professionals who need answers that are verifiably correct — not just plausible.
- [Quick Start](https://docs.mira-app.dev/get-started/quick-start.md): Install MIRA, connect a provider, and get your first answer in under 5 minutes.
- [Error Reference](https://docs.mira-app.dev/reference/error-reference.md): Common MIRA error messages and how to resolve them.
- [Glossary](https://docs.mira-app.dev/reference/glossary.md): Definitions of key terms used throughout the MIRA documentation.
- [Settings Reference](https://docs.mira-app.dev/reference/settings-reference.md): Complete reference for every setting available in the MIRA Settings panel.

## OpenAPI Specs

- [openapi](https://docs.mira-app.dev/api-reference/openapi.json)