awesome-cli-coding-agents

Curated directory of terminal-native AI coding agents and the harnesses that orchestrate them. Covers open-source tools (Pi, OpenCode, Aider, Goose), platform agents (Claude Code, Codex, Gemini CLI), parallel runners, autonomous loops, and agent infrastructure.
https://github.com/bradagi/awesome-cli-coding-agents

Last synced: 13 days ago
JSON representation

Harnesses & orchestration
- Agent infrastructure
  - claude-code-router
  - agent-browser
  - NemoClaw - level security policies via OpenShell runtime. Apache-2.0.
  - OpenWork - source alternative to Claude Cowork for teams; local-first desktop app powered by OpenCode with one-click setup. MIT.
  - open-claude-cowork - source version of Claude Cowork with 500+ SaaS app integrations.
  - OneCLI - source credential vault for AI agents; Rust HTTP gateway injects API keys transparently so agents never handle raw secrets. Per-agent scoped tokens, AES-256-GCM encryption at rest. Apache-2.0.
  - Claude Code Tools
  - Coasts
  - subtask
  - claude-cmd
  - toprank - LD schema, keyword bids) directly to source code or CMS. Open-source, MIT.
  - kasetto
  - claudebox
  - skill-optimizer
  - Wit - sitter AST parsing; agents declare intents, acquire symbol-level locks, and get conflict warnings before writing code. JSON-RPC daemon. MIT.
  - Untether
  - brood-box - isolated microVM sandbox for AI coding agents (Claude Code, Codex, OpenCode) with COW snapshot isolation, egress control, and MCP authorization.
  - Nex - tool MCP server (`npx @nex-ai/nex`) and persistent memory across agent sessions. MIT.
  - AgentManager
  - AgentLint - backed checks for AI-friendly repos. Scans file structure, instruction quality, build setup, session continuity, and security posture. Claude Code plugin with auto-fix. Your AI agent is only as good as your repo.
  - EchoCoding - triggered SFX, ambient soundscape, and optional cloud TTS/ASR voice interaction for Codex and Claude Code workflows.
  - agent-terminal - pty; capture output and send input programmatically.
  - zosma-qa
  - gate4agent - RPC 2.0) modes with tokio broadcast fan-out. MIT.
  - Praman - driven test generation via MCP. Apache-2.0.
  - lifeos-cli - native LifeOS for managing intentions, habits, and timelogs; provides structured persistent context for agentic workflows.
  - Agentic Engineering Framework - neutral governance framework for CLI coding agents. Structural enforcement of task-driven workflows, context budget management, antifragile healing loops, and audit compliance. Works with Claude Code, Aider, Cursor, and any file-based agent.
  - OSOP - editor.vercel.app) and [spec](https://github.com/Archie0125/osop-spec).
  - claude-northstar
  - AgentPlane - native workflow (task → plan → approve → implement → verify → finish). All state stays in `.agentplane/` inside the repo; no hosted runtime. MIT.
  - linear-cli - file Python CLI for Linear (the issue tracker), zero dependencies. Designed for use as a subagent tool by Claude Code, Codex, Gemini, or Cursor; ships a SKILL.md for drop-in Claude Code integration. MIT.
  - Not Human Search - plugin); 8,000+ indexed sites exposed via MCP server, REST API, and full-text search. Lets agents discover and verify external services before wiring them into a repo. MIT.
  - agent-lsp - aware language intelligence: 50+ LSP tools, speculative execution (preview edits without touching disk), 21 skills, 30 CI-verified languages. Measured 5–34× token savings vs grep/read. Works with Claude Code, Codex, Gemini CLI, Goose, and any MCP client. Go, MIT.
  - Agent FM - 2.0.
  - AgentTier - native sandbox runtime for AI coding agents. A `Sandbox` CRD provisions a Pod + PVC + NetworkPolicy with optional gVisor isolation; the `agenttier` Go CLI runs agent invocations that stream stdout/stderr/exit as SSE. Ships reference templates for Claude Code + Bedrock and LangGraph. Apache-2.0.
  - Camofox Browser - compatible with anti-detection, human-like fingerprinting, and a REST API for agent tool integration. MIT.
  - AgentSight - instrumentation eBPF observability for LLM/coding agents; captures syscall-level traces (file, process, network) without modifying the agent. MIT.
  - Harness Starter Kit - first harness engineering starter kit for adding durable agent instructions, drift checks, failure memory, and evaluation loops to repositories. MIT.
  - agenttrace - first Bubble Tea TUI for inspecting Claude Code, Codex CLI, Gemini CLI, Aider, Cursor, Hermes, OpenCode, Kimi, and Copilot-style session logs; surfaces cost, cache usage, failures, latency, anomalies, health gates, and diffs.
  - pi-mem - Markdown persistent memory for coding agents; long-term, daily, scratchpad, and searchable notes with zero dependencies on vector DBs. MIT.
  - authsome
  - ActPlane - level agent harness that compiles a policy DSL to an eBPF engine for labeled information-flow control at the syscall boundary, with corrective feedback. MIT.
  - pi-reflect - improving behavioral files for coding agents; automated self-reviews that evolve AGENTS.md rules from actual mistakes. MIT.
  - AgentDiff - native provenance for AI-written code: hooks every CLI agent (Claude Code, Codex, Cursor, Gemini, OpenCode, Windsurf, Copilot) and records which agent wrote which line, reconciles it against each commit, and signs every attribution with ed25519. Records live in your own git refs (`agentdiff list/blame/report`); no server. Rust, MIT/Apache-2.0.
  - codex-profiles - profile`. MIT, no token copying.
  - Agent Memory System - agent workflows. MIT.
  - machine
  - grite - backed issue tracker with CRDT merging for AI coding agents. Issues live as an append-only event log in `refs/grite/wal`, sync via `git push`/`fetch`, and converge deterministically across agents — no server, no database, no merge conflicts. Stable `--json` output and a `grite install-skill` command for Claude Code. MIT.
  - pi-builder - agent, Crush, gptme) behind a single interface; capability-based routing, health caching, fallback chains, SQLite persistence, and a streaming OrchestratorService. MIT.
- Orchestrators & autonomous loops
  - claude-flow - agent swarms with coordinated workflows.
  - gastown - agent orchestration with persistent work tracking.
  - ralph-orchestrator - based system maintaining agents in a loop until task completion.
  - ralph-tui
  - AgentsMesh - agent collaboration via channels and pod bindings, built-in Kanban with MR/PR integration. Self-hosted with BYOK. Supports Claude Code, Codex CLI, Gemini CLI, Aider, OpenCode. BSL-1.1.
  - loom - agent coordination.
  - Bernstein - commits.
  - wreckit
  - kodo - agent coding orchestrator that directs Claude Code, Codex, and Gemini CLI through work cycles with independent architect and tester verification. SWE-bench verified.
  - OpenCastle - agent orchestration framework that turns AI coding assistants (Copilot, Cursor, Claude Code, OpenCode, Windsurf, Codex CLI) into 19 coordinated specialist agents. CLI-driven (`npx opencastle init`), with task decomposition, parallel work, and quality gates. MIT.
  - ORCH - retry, inter-agent messaging, and TUI dashboard.
  - Forge - driven development loop for Claude Code; three-command pipeline (brainstorm specs, plan task DAGs, execute autonomously) with context survival, backpropagation that traces bugs to spec gaps, and Claude-on-Claude code review. MIT.
  - agx - based execution engine for AI coding agents; durable Wake→Work→Sleep loops that resume instantly across sessions. Supports Claude Code, Codex CLI, Gemini CLI, and Ollama. CLI + web dashboard + macOS app.
  - sage - agnostic support (Claude Code, Cline, Codex, Gemini CLI, ACP), wave-based plan execution, git worktree isolation, MCP integration, skills system, headless CI mode, and 295 bats tests. MIT.
  - Aeon - healing loop. MCP + A2A integrations. MIT.
  - Relay - first orchestrator that runs inside your existing Claude or Codex CLI via MCP; classifies a request, decomposes it into tickets with a dependency DAG, dispatches across one or more repos, and supervises with live PR tracking + approval gates. CLI, TUI (ratatui), and GUI (Tauri) dashboards share `~/.relay/` state. MIT.
  - great_cto - management layer of 34 specialist AI agents covering the full SDLC (architect, PM, senior-dev, reviewer, QA, security, devops, L3-support + 18 archetype-specific reviewers) with auto-detected archetypes and compliance gates (PCI-DSS, HIPAA, FedRAMP, GDPR, EU AI Act). Runs in Claude Code, Cursor, Codex CLI, Aider, and Continue via AGENTS.md + MCP. MIT.
  - The Factory - evolving meta-harness for autonomous software dev and research; turns any codebase into an auto-research project, auto-discovers eval dimensions, generates scoring harness, and runs keep/revert experiment loops with monotonic-improvement guards. Multi-contributor. MIT.
  - Galley - first runtime for supervised AI coding tasks: isolated git worktrees, supervisor review against acceptance criteria, retry/escalate loops, on-disk run evidence, and PR handoff. Supports Codex CLI and Claude Code. Go, MIT.
  - Loki Mode - to-product autonomous loop with a built-in verification gate: a reason/act/reflect/verify closure plus a blind-review completion council that can veto "done", so it will not mark work complete until the evidence passes. Brownfield healing (`loki heal`), local-first BYO-keys, 26-tool MCP server, reads AGENTS.md. Source-available (BUSL-1.1).
  - OMK - neutral CLI control plane for coding agents: routes runtimes, scopes MCP, runs DAG workers, and verifies evidence before completion. MIT.
  - Hephaestus - agent builder, A2A Hub routing, local ontology, and memory/security gates. Apache-2.0.
- Session managers & parallel runners
  - vibe-kanban
  - cmux - source platform for running multiple coding agents in parallel.
  - Superset
  - Claude Squad - based harness to run and manage multiple Claude Code sessions side-by-side.
  - Emdash
  - CodexMonitor
  - Crystal
  - Toad
  - mux
  - supacode
  - jean
  - Parallel Code
  - CLI Agent Orchestrator (CAO) - agent orchestration via tmux with intelligent task delegation patterns.
  - Catnip
  - vibe-tree
  - amux - healing watchdog, kanban board, agent-to-agent REST API, and mobile PWA. Single Python file, Python 3 + tmux. MIT.
  - AgentPipe - agent conversations by enabling different AI CLI tools (Claude Code, Gemini, Qwen, etc.) to communicate in shared rooms. MIT.
  - amux
  - CliDeck - like browser dashboard for managing multiple CLI coding agents (Claude Code, Codex, Gemini CLI, OpenCode) with live status detection, session resume, autopilot routing, and full control from a phone while away. MIT.
  - multi-agent-workflow-kit
  - PATAPIM - terminal grid for running multiple CLI coding agents simultaneously; features AI state detection, built-in Whisper voice dictation, LAN remote access, and an embedded MCP browser. Built with Electron and node-pty. Freemium.
  - CLITrigger - hosted web UI for orchestrating Claude Code, Codex, and Gemini CLIs in parallel git worktrees. Features multi-agent discussion mode (architect/developer/reviewer debate before implementation), cross-project Morning Review Queue, scheduled execution with rate-limit auto-recovery, and a built-in Git client. MIT.
  - agent-deck - aware, MCP integration, 8+ contributors. MIT.
  - agent-of-empires - friendly). Rust, uses tmux + git worktrees. MIT.
  - ntm
  - Cate - pty. MIT.
  - Agent Sessions - first macOS session-history browser for AI coding agents, with transcript search across Codex, Claude Code, OpenCode, Cursor Agent, Hermes, Copilot CLI, OpenClaw, and more; resume is available where the underlying CLI supports it. MIT.
  - hcom - turn without changing how you use them. TUI dashboard, collision detection, cross-device relay. Rust, MIT.
  - Clave - agent CLIs (Claude Code, Gemini CLI, Codex) in parallel — split/grid terminal layouts, per-project session groups, a built-in git panel, and remote sessions over SSH. Fully local, no account. Electron. MIT.
  - tmuxlet - p` style blocking interface. Single binary, zero deps. Works against the regular Claude subscription bucket (not the separate Agent SDK credit) by driving interactive Claude Code from the outside.
Terminal-native coding agents
- Closed Source
  - Claude Code - aware terminal agent for code edits, refactors, and git workflows. Source-available, no OSS license.
  - Warp - in AI agent mode; understands tasks, runs commands, edits files, and orchestrates multi-step workflows.
  - GitHub Copilot in the CLI
  - Droid - model CLI coding agent; #1 on Terminal-Bench, specialized droids for different tasks, headless CI mode, and multi-interface support (CLI/IDE/Slack/Linear).
  - Command Code - 1 neuro-symbolic AI; adapts to preferences over time with project-specific taste profiles.
  - FetchCoder
  - Amp
  - Devin - healing code, and PR collaboration.
  - Cursor CLI - line agent (`agent`) with shell mode, headless/CI support, parallel agents, and multi-model access.
  - Tabnine CLI - powered terminal coding assistant with agentic workflows; distributed as a Docker container, requires enterprise license.
  - Mentat CLI - native coding agent CLI for managing remote Mentat agents from your terminal; auto-detects repo/branch context.
  - Cortex Code CLI - native AI coding agent CLI for building pipelines, analytics, and AI apps with enterprise governance.
- OpenClaw ecosystem
  - OpenClaw - source personal AI assistant. Sorted by GitHub stars.
  - nanobot - lightweight ~4,000-line Python rewrite of OpenClaw; tool use, persistent memory, scheduled tasks, and multi-channel support (Telegram/Discord/WhatsApp). MIT.
  - ZeroClaw - driven pluggable architecture (providers, tools, memory, channels), runs on minimal hardware (<5MB RAM), multi-channel CLI/Telegram/Discord/Slack, with sandboxed execution and hybrid vector+keyword search.
  - PicoClaw - lightweight personal AI assistant in Go inspired by OpenClaw; runs on $10 hardware with less than 10MB RAM.
  - IronClaw - based permissions, and prompt injection defense.
  - NullClaw - compatible agent in Zig; 678KB static binary, ~1MB RAM, <2ms startup, 23+ providers, 18 channels. MIT.
  - Clawith - agent collaboration platform that scales OpenClaw to organizations. Apache-2.0.
  - Moltis - native alternative to OpenClaw; zero unsafe code, 2,300+ tests, built-in voice I/O, MCP servers, and multi-channel access. MIT.
  - GitClaw - native AI agent framework where agent identity, rules, memory, tools, and skills are all version-controlled files. MIT.
  - LionClaw - first local AI CLI with a small auditable kernel, durable sessions, and installable skills. MIT.
  - NanoClaw - first lightweight alternative to OpenClaw; runs agents in Apple containers/Docker with sandboxed execution, built on Anthropic's Agents SDK.
  - claw0 - to-1 tutorial companion for the OpenClaw ecosystem; walks through building an agent harness from scratch, covering planning, context compression, and task persistence.
- Open Source
  - OpenCode - native coding agent with 75+ provider support, LSP integration, and privacy-first design (formerly opencode-ai; now at opencode.ai).
  - Hermes Agent - improving CLI agent with persistent memory, automated skill creation, sandboxed code execution via Unix socket RPC, and multi-platform reach (Telegram/Slack/Discord/WhatsApp); supports 300+ models across multiple providers.
  - Gemini CLI - 2.0.
  - Codex CLI - 2.0.
  - OpenHands - source agentic developer environment (formerly OpenDevin) with CLI and web entrypoints; also has a lightweight [CLI-only package](https://github.com/OpenHands/OpenHands-CLI).
  - Open Interpreter
  - Cline CLI - agnostic autonomous agent for planning, file edits, command execution, and browser use.
  - Aider - programming agent for editing files via diffs/patches, with strong git and multi-file workflows.
  - Goose - device and integrate with MCP.
  - Pi - mono toolkit; unified LLM API, TUI, skills, and MCP support.
  - Continue CLI - source terminal extension for multi-model coding with local/privacy focus.
  - Qwen Code - 2.0.
  - Crush - provider, LSP-aware, with rich terminal UI.
  - Roo Code CLI - mode CLI agent (architect/code/debug/orchestrator modes); Claude-like terminal interface with skills and checkpoints.
  - SWE-agent - bench-style workflows.
  - Kilo Code CLI
  - Plandex - first" CLI agent for building features across multiple files with structured steps and 2M token context.
  - Smol Developer
  - Trae Agent - friendly CLI agent for software engineering tasks, with modular architecture and multi-LLM support. MIT.
  - Claude Engineer - driven CLI for agentic Claude workflows with file management and iterative development.
  - Claurst
  - Kimi CLI - 2.0.
  - Free Code
  - ForgeCode - scale refactor tooling.
  - Kode CLI - source CLI agent for terminal-native coding with multi-provider support.
  - Codebuff - agent AI coding assistant with CLI support for collaborative coding workflows.
  - gptme - lived self-modifying agents with git-backed memory via [gptme-agent-template](https://github.com/gptme/gptme-agent-template).
  - Mistral Vibe - 2.0.
  - Every Code - agents, theming, and multi-provider orchestration (OpenAI, Claude, Gemini). Apache-2.0.
  - Devon - source pair programmer with a TUI; autonomous planning, execution, and debugging in Git workflows.
  - OH-MY-PI - first ergonomics.
  - AutoCodeRover
  - Grok CLI - based coding tasks.
  - Codel
  - CodeMachine-CLI - agent CLI aimed at running coding workflows locally (vibe-coding oriented).
  - Letta Code - first CLI coding agent built on the Letta platform (formerly MemGPT); persistent memory across sessions, model-agnostic (Claude/GPT/Gemini), skill learning, and context repositories.
  - RA.Aid - full autonomy.
  - open-codex - provider support (OpenAI, Gemini, OpenRouter, Ollama).
  - Agentless
  - Amazon Q Developer CLI - 2.0.
  - Neovate Code - model/multi-provider support, MCP integrations, and headless automation mode. MIT.
  - Groq Code CLI - fast inference; extensible tools/commands with multi-model support.
  - Dexto - ready coding agent with sub-agent spawning.
  - g3 - running, repo interaction, skills system, and provider abstraction.
  - Coro Code - source CLI coding agent, a free alternative to Claude Code; generate, debug, and manage code seamlessly.
  - LettaBot
  - Mini-Kode
  - nori-cli - provider CLI built on Codex CLI; switch between Claude, Gemini, and Codex from the same native terminal.
  - cursor-agent - based agent replicating Cursor's coding assistant capabilities; supports Claude, OpenAI, and local Ollama models.
  - VibePod - config setup, local metrics, HTTP traffic tracking, and an analytics dashboard for side-by-side comparison.
  - QQCode
  - picocode - based coding agent focused on CI workflows and small codemods; multi-LLM with personas.
  - Smelt - provider (Anthropic, OpenAI, Ollama, GitHub Copilot, any OpenAI-compatible endpoint), four modes (Normal/Plan/Apply/Yolo), granular permission system, parallel subagents, vim keybindings, and headless scriptable mode. MIT.
  - Binharic - provider "tech-priest persona" coding agent CLI (stylized, tool-using).
  - Darce - model CLI agent built with Ink; 7 tools, smart model routing across providers, streaming, session resume, and slash commands. MIT.
  - CLAII - first AI coding agent with multi-agent orchestration, MCP toolchains, and memory-persistent refactors.
  - Claw Code - room Python/Rust rewrite of Claude Code architecture using oh-my-codex; fastest repo in GitHub history to 100K stars. Born from the March 2026 Claude Code source leak. MIT.
  - claw-code-agent - only Claude Code rewrite with zero external dependencies; born from the March 2026 Claude Code source leak, positioned as easier to hack on than the Rust/TypeScript reimplementations.
  - openHarness - source Claude Code alternative. 78 slash commands, 42 tools, MCP (stdio/HTTP/SSE + OAuth 2.1), hooks, subagents, plan mode. Works with Anthropic/OpenAI/Ollama/llama.cpp/LM Studio. Ships both npm and Python SDK. MIT.
  - Crab Code - native agentic coding CLI aligned with Claude Code's toolset and permission model; works with any LLM provider (Anthropic/OpenAI/DeepSeek/Bedrock/Vertex). 27 crates, 4,700+ tests, layered config system. Apache-2.0.
  - Octomind - source, model-agnostic AI agent runtime with community tap registry (`developer:rust`, `doctor:blood`, `legal:contracts`), MCP support with runtime self-extension, 13+ providers, and adaptive compression. Written in Rust. Apache-2.0.
  - Nanocoder - first CLI coding agent built by a community collective; bring your own model (Ollama, OpenRouter, or any OpenAI-compatible API), native tool calling with XML fallback, MCP support, and file-based custom commands and tools. MIT.
  - VT Code - source coding agent with LLM-native code understanding and robust shell safety. Supports multiple LLM providers with automatic failover and efficient context management. MIT.
  - zot - overhead and lightweight coding agent harness with TUI/JSON/RPC modes, structured tools, reviewable file diffs, skills, extensions, and optional guardrails.
  - San - native runtime for specialized AI agents; provider-agnostic (Anthropic, OpenAI, Google, DeepSeek, Moonshot, Qwen, MiniMax, GLM), runs Claude Code skills/plugins/MCP unmodified, swappable search backends, custom personas, sandboxed subagents, lifecycle hooks, and a self-evolving memory loop. Single ~12 MB binary, zero runtime deps. Apache-2.0.
  - Keen Code - based CLI coding agent focused on efficient context management — uses lean TurnMemory summaries instead of raw tool traces to maximize context efficiency. 9+ providers, skill-driven MCP servers, and full transparency with every prompt, design doc, and implementation plan saved as markdown. MIT.
  - Zap - first Rust TUI coding agent that injects only the context your task needs — no system prompt bloat. Single binary, no runtime. Supports Claude, Gemini, OpenAI, and local models via LM Studio; code-indexed via SQLite for fast symbol lookup; MCP support. MIT.
  - DvalinCode - neutral coding agent with three modes (Chat/Cowork/Code), inline diff approval, built-in Web GUI launched from a single binary. Works with any OpenAI-compatible endpoint (DeepSeek, OpenAI, Claude via OpenRouter, Groq, Ollama); macOS shell sandbox via `sandbox-exec`; `.dvalincodeignore` for sensitive files; `AGENTS.md` project memory. MIT.

Programming Languages

TypeScript 43 Python 24 Rust 19 Go 18 JavaScript 6 Shell 4 C 2 Swift 2 Zig 1

ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Awesome

awesome-cli-coding-agents

Harnesses & orchestration

Agent infrastructure

Orchestrators & autonomous loops

Session managers & parallel runners

Terminal-native coding agents

Closed Source

OpenClaw ecosystem

Open Source