Amari: A Cyberpunk-Style Local-First AI Agent Terminal

Amari is a local-first AI agent terminal built with Next.js, featuring a cyberpunk visual style. It supports multi-platform LLMs including Claude, Gemini, Kimi, OpenAI, OpenRouter, and local Ollama, has intelligent tool loop capabilities, and works on both macOS and Windows. The project is maintained by Antastik and hosted on GitHub (project name: amari, link: https://github.com/Antastik/amari), released on June 15, 2026.

With the popularity of AI tools today, developers and advanced users often need to switch between platforms like Claude (deep reasoning), Gemini (multimodal), OpenAI (general conversation), and Ollama (local privacy), leading to low efficiency and difficulty in automating complex workflows. Amari was created to address this pain point, integrating multi-platform access, intelligent tool calls, and workflow orchestration into a unified interface.

Core Features

1. Unified Access to Multi-Platform LLMs: Supports Claude (deep reasoning), Gemini (multimodal), Kimi (long context), OpenAI (general purpose), OpenRouter (gateway), and Ollama (local privacy).
2. Local-First Architecture: Local data storage, offline availability, and privacy protection.
3. Intelligent Tool Loop: Processes complex tasks through the flow: user input → model reasoning → tool call → result feedback → continued reasoning → final output.
4. Cyberpunk Style: Neon colors, terminal interface, dynamic effects, frosted glass effects, and geometric elements.

Technical Architecture

• Frontend framework: React + Next.js (server-side rendering, API routing, TypeScript support)
• Styling: Tailwind CSS + custom CSS
• Cross-platform: Supports macOS/Windows desktop apps, web apps, and Vercel deployment.

Use Cases

1. Multi-Model Comparison: Quickly switch models to compare answer quality.
2. Local Privacy Protection: Use Ollama to process sensitive documents without uploading data to the cloud.
3. Complex Task Automation: Automate multi-step tasks like searching for papers → summarizing → generating PPT outlines.
4. Development Assistance: Code review, document querying, API debugging, and error analysis.

Comparison with Similar Tools

Feature	Amari	ChatGPT	Claude Desktop	Ollama WebUI
Multi-platform support	✅ 6+ platforms	❌ OpenAI only	❌ Claude only	❌ Local models only
Local-first	✅ Yes	❌ No	⚠️ Partial	✅ Yes
Agent tools	✅ Built-in	⚠️ Requires plugins	✅ Supported	⚠️ Limited
Cyberpunk style	✅ Unique	❌ Standard	❌ Standard	❌ Standard
Open source	✅ Yes	❌ No	❌ No	✅ Yes

Current Limitations

• Tool ecosystem: Built-in tools are fewer than mature frameworks like LangChain.
• Mobile support: Desktop-focused; mobile experience needs optimization.
• Collaboration features: Lack of team collaboration and conversation sharing.

Future Directions

• Plugin system: Open API for community tool integration.
• Workflow orchestration: Visual editor for complex task orchestration.
• Knowledge base integration: Support RAG to connect local knowledge bases.
• Voice interaction: Integrate speech recognition and synthesis.

Amari represents the development direction of AI terminal tools: unified multi-platform access, local-first approach, intelligent agent, and unique experience. The cyberpunk style is an interpretation of human-machine relationships in the AI era. It is suitable for users who frequently use multiple models, value privacy, and pursue efficient workflows. Its open-source nature and unique design give it a competitive edge in the field.