Agent runtimes you install and use. Coding-agent CLIs (Claude Code, Codex, Cursor, Gemini CLI, Aider), standalone agent runtimes (Hermes, OpenClaw, OpenHarness), and the meta-harnesses that manage sessions, quotas, and multi-agent workflows on top of them.
16 harnesses
OpenAI's coding-agent CLI. Runs Codex models with tool use, shell execution, and file editing. Ships with the Responses API integration so context management happens inside the model call.
Cursor's coding-agent runtime built into the Cursor IDE. Reads .cursorrules, MCP servers, and Cursor commands. Available as CLI (cursor-agent) and inside the editor.
AI pair programming in your terminal. Model-agnostic harness that speaks to Claude, GPT, Gemini, and local models. Best-in-class git integration with auto-commits per edit.
Multi-harness control plane for Claude Code, Codex, Cursor, and OpenCode. Quota-aware rotation across multiple subscriptions, shared thread state, and local-first execution.
Open-source coding-agent harness with model-agnostic backend. Terminal-first, supports Claude, GPT, and local models. MCP-native.
Agent harness for complex codebases. Wraps Codex, OpenCode, and other coding-agent CLIs with token-efficient context management for long-running work.
Autonomous coding-agent harness that runs as a VS Code extension. Reads cline_rules, supports MCP, and executes multi-step tasks with human-in-the-loop approval.
Mission Control for Claude Code. Native desktop app to browse, search, and drive Claude Code sessions with live status and diff review.
Terminal UI to manage AI coding-agent sessions (Claude Code, OpenCode, Codex, Grok, Gemini CLI) in tmux. Live status, group tree, quick prompts, diff review.
Open-source visual workspace and harness for Claude Code and Codex. Browser-based UI over the terminal harness with session sharing and review tooling.
Google's coding-agent CLI running Gemini models. Terminal-based, supports MCP servers, and integrates with the Google Cloud agent stack.
Nous Research's open-source autonomous AI agent. Runs as a persistent daemon on your own infrastructure, connects to 16+ messaging platforms (Telegram, Discord, Slack, WhatsApp), and writes its own reusable skills from experience. Model-agnostic — Anthropic, OpenAI, Google, DeepSeek, Ollama.
Open-source AI agent framework with a complete runtime, CLI, backend services, and web dashboard. Multi-model support (GPT-4o, Claude, Gemini, DeepSeek) with native Telegram/Discord/WhatsApp integration. Emerged as a Claude Code alternative after Anthropic paywalled third-party tool usage in April 2026.
Ultra-lightweight pure-Python alternative to Claude Code from HKU Data Intelligence Lab. ~98% of Claude Code's core tool capabilities in only 11,733 lines of code. Fully compatible with Anthropic Skills and the Claude Code plugin ecosystem. Supports any OpenAI-compatible model.
OS-like agent runtime that sits between the model and your application. LobeHub's Chief Agent Operator hires agents from a 273K skill marketplace and runs them 24/7 in the cloud, reporting through Slack, Discord, and other platforms. Model as CPU, context window as RAM.
Anthropic's official coding-agent CLI. Runs Claude in your terminal with tool use, file operations, and shell access. The reference harness that skills, sub-agents, hooks, and slash commands are built for.