OpenAnalyst CLI
The Universal AI Agent for Your Terminal. One CLI, every LLM provider, persistent conversations across models.
Quick Install
Single command install on macOS, Linux, and Windows. No runtime dependencies.
7 Providers
Claude, GPT, Gemini, Grok, OpenRouter, Bedrock — switch mid-conversation.
Full Terminal UI
Full-screen TUI with scrollable chat, sidebar, vim-mode, and modal dialogs.
Multi-Agent
Orchestrate parallel agents, autonomous loops, and swarm task decomposition.
Overview
OpenAnalyst CLI is an independent, open-source AI coding agent that connects to every major LLM provider through a single, unified terminal interface. It features a full full-screen terminal UI, multi-provider streaming, multi-agent orchestration, 51+ slash commands, 19 built-in tools, and MCP protocol support.
Why OpenAnalyst?
| Capability | OpenAnalyst CLI | Other CLI Tools |
|---|---|---|
| Providers | 7 first-class (OpenAnalyst, Anthropic, OpenAI, xAI, Gemini, OpenRouter, Bedrock) | Typically 1–2 |
| Mid-conversation switching | Session persists across providers | Not supported |
| Terminal UI | Full Full-screen TUI (default) | Basic REPL |
| Multi-agent | Built-in parallel orchestration | Limited or none |
| Multimedia | /image, /voice, /speak, /vision, /diagram | Rarely supported |
| Slash commands | 51+ | 5–15 typical |
| Binary | Native binary, zero runtime deps | Often needs Node/Python |
Key Features
Universal Provider Support
Connect to OpenAnalyst, Anthropic Claude, OpenAI GPT, Google Gemini, xAI Grok, OpenRouter (350+ models), and Amazon Bedrock — all with live model discovery and streaming.
Full Terminal UI
Scrollable chat with inline tool cards, startup banner, status line with spinner and token counts, vim-mode input, permission dialogs, and mouse navigation.
Multi-Agent Orchestration
Spawn sub-agents for parallel tasks, run autonomous agent loops with think→act→observe→verify cycles, and dispatch Mixture of Experts for complex problems.
51+ Slash Commands & 19 Tools
Git operations, multimedia generation, web scraping, planning, analysis, Playwright browser control, and more — all from your terminal.
Permission System & Sandboxing
Three permission modes (read-only, workspace-write, full-access) with modal approval dialogs and filesystem sandboxing.
Single Native Binary
Native binary with zero runtime dependencies. Fast startup, low memory, and works on macOS, Linux, and Windows.
Documentation Pages
Getting Started
Installation →
Install on macOS, Linux, or Windows. Build from source.
Quick Start →
Login, launch the TUI, and run your first prompt in 2 minutes.
Authentication →
OAuth login, API keys, credentials file, and provider resolution.
Core Concepts
LLM Providers →
7 first-class providers, model aliases, mid-conversation switching, feature matrix.
Terminal UI →
TUI layout, keybindings, vim mode, tool call cards, voice input, agent panel.
Slash Commands →
51+ commands for session, git, multimedia, web scraping, analysis, and agents.
Built-in Tools →
19 tools: file ops, search, web, bash, PowerShell, notebooks, and more.
Advanced
Multi-Agent System →
Sub-agents, swarms, autonomous loops, MOE, model routing, effort budgets.
Configuration →
Settings hierarchy, permissions, hooks, environment variables, file structure.
MCP Integration →
Model Context Protocol: transports, server config, tool discovery, popular servers.
Internal architecture and module organization.