$ find your agent harness

MIT-licensed self-hosted personal AI assistant ('the lobster way') that exploded to ~382k GitHub stars in months: any model, 20+ messaging channels (WhatsApp/Telegram/Slack/iMessage...), SKILL.md skills via ClawHub, Docker-sandboxed sessions, cron jobs, voice, and a live Canvas.

Nous Research's MIT-licensed 'self-improving' server-resident agent (Feb 2026, ~211k stars): model-agnostic with persistent cross-session memory, autonomous skill creation compatible with the agentskills.io standard, DSPy/GEPA evolutionary self-optimization, and execution backends from local/Docker to Modal and Daytona.

The clean-room Rust/Python rewrite of Claude Code's harness architecture born hours after the March 31, 2026 source leak — reportedly the fastest-growing repo in GitHub history (~195k stars) with permission-gated tools, MCP, agent swarms, and container-first deployment — though the project is chaotic, largely agent-maintained, and its details vary across sources.

Anthropic's Claude-only agent harness in native desktop form, redesigned April 14, 2026 around a multi-session sidebar for parallel agents across repos, with integrated terminal/editor/diff viewer, cloud-hosted 'Routines' that run on Anthropic infrastructure while your machine is off, and the ecosystem that originated Agent Skills, subagents, hooks, plugins, and the headless Agent SDK (repo is issue tracker/docs; stars 136,750).

Anthropic's terminal-native agentic coding tool and the category-defining harness, distinguished by its layered extensibility model (CLAUDE.md memory, Agent Skills, subagents, hooks, plugins, MCP) plus permission modes, checkpoints, and optional sandboxing, available from CLI/IDE/desktop/web with an Agent SDK for building on top.

Google's fully open-source (Apache-2.0) terminal agent with a generous free tier, distinguished by built-in Google Search grounding, an extensions ecosystem bundling prompts/MCP servers/commands, GEMINI.md context files, checkpointing, and real-time voice mode.

MIT-licensed Python library (103k stars) that lets any LLM drive a real Chrome browser via DOM+vision instead of brittle selectors, now flanked by a desktop app and a usage-based cloud with stealth browsers, proxies, and a 24/7 'Browser Use Box' agent.

OpenAI's open-source (Apache-2.0) Rust-rewritten terminal coding agent, distinguished by kernel-level OS sandboxing (Apple Seatbelt on macOS, Landlock/seccomp on Linux), a 272K-1M token context with GPT-5.x Codex models, MCP support, and cloud handoff to Codex web tasks.

High-performance Rust editor (GPL, hit 1.0 in 2026) whose Agent Panel offers a native agent with profiles/tool-permissions and MCP, but whose real differentiator is the open Agent Client Protocol (ACP) that lets external agents (Claude Code, Codex CLI, Gemini CLI, OpenCode — which bring their own skills support) render diffs natively in-editor, plus first-in-class native Parallel Agents (April 2026).

Formerly OpenDevin, the leading open-source autonomous software-engineering platform: Docker-sandboxed agents that write code, run commands, browse, and open PRs, with a production agent SDK hitting 72% on SWE-bench Verified (Claude Sonnet 4.5) and its own broader OpenHands Index benchmark launched Jan 2026.

Mario Zechner's radically minimal MIT/TypeScript harness (~68k stars): four tools (read/write/edit/bash), a ~300-word system prompt, deliberately no built-in MCP/subagents/permission-popups, with everything else added via TypeScript extensions, skills, and npm/git-shareable packages across interactive, print/JSON, RPC, and SDK modes.

Warp is a Rust-based GPU-accelerated 'Agentic Development Environment' (terminal app, source opened April 2026 under dual MIT/AGPL) distinguished by running multiple agents in parallel — including hosting Claude Code, Codex, and Gemini CLI — plus Oz, its cloud orchestrator for autonomous background agents, with BYOK and credit-based pricing.

MIT-licensed Python framework (~55k stars) for role-based multi-agent 'Crews' plus event-driven 'Flows', with unified memory, hierarchical delegation, and a commercial AMP cloud platform (free tier, Pro from ~$25/mo, enterprise) that CrewAI says powered ~2B agentic executions in the year to January 2026.

Block's Rust-built, Apache-2.0 general engineering agent — now stewarded by the Linux Foundation's Agentic AI Foundation — distinguished by the deepest MCP integration in the ecosystem (70+ extensions, can also act as an MCP server) and shareable YAML 'Recipes' that scaled agent workflows to 60% of Block.

Low-level graph-based agent orchestration runtime with automatic checkpointing, durable execution, human-in-the-loop interrupts, and streaming (v1.2 in May 2026 added typed streaming, per-node timeouts, and distributed runtime), used in production by Klarna, LinkedIn, Uber, and Replit and serving as the substrate under Deep Agents.

Hugging Face's minimalist (~1,000-line core, Apache 2.0) 'agents that think in code' harness where the model writes executable Python as its action format instead of JSON tool calls, model-agnostic via transformers/Ollama/LiteLLM, with optional sandboxed execution (E2B, Docker, WebAssembly), managed sub-agents, and Hub sharing of tools and agents.

The original open deep-research harness (Apache-2.0, 28k stars): a planner/executor multi-agent pipeline that fans out parallel web and local-document searches into cited reports, with recursive tree-like 'Deep Research', MCP data-source integration, and any-LLM support.

OpenAI's agent harness whose April 15, 2026 overhaul added a 'model-native harness' with Codex-like filesystem/shell tools, built-in compaction and memory, snapshot/rehydrate durable execution, handoffs, guardrails, tracing, and native sandbox execution via bring-your-own or built-in providers (E2B, Modal, Daytona, Cloudflare, Vercel, Blaxel, Runloop); non-OpenAI models supported via LiteLLM and Chat Completions-compatible endpoints.

The leading TypeScript agent framework (Apache 2.0 core since July 2025; some enterprise features commercial) offering agents, type-safe workflows with suspend/resume human-in-the-loop, memory, a unified router to 3,300+ models across ~94 providers, guardrails/scorers/tracing, Mastra Studio for interactive testing, and file-based agent definitions added July 2026.

LangChain's self-described 'batteries-included agent harness' (GA March 2026, ~26k stars) for long-running tasks, layering planning via write_todos, a virtual filesystem for context offloading, SKILL.md-compatible skills with progressive disclosure, subagent spawning, and LangGraph-store-backed long-term memory on top of the LangGraph runtime.

Alibaba's Apache-2.0 terminal agent (originally a Gemini CLI fork) optimized for Qwen3-Coder models, distinguished by broad provider flexibility (OpenAI/Anthropic/Gemini/Qwen APIs plus Ollama/vLLM local models), skills/subagents primitives, headless mode, and TypeScript/Python/Java SDKs.

Kilo's open-source terminal agent (Kilo CLI 1.0, built on the MIT-licensed OpenCode foundation) distinguished by support for 500+ models with zero-markup BYOK pricing, an MCP marketplace, and cross-surface continuity where a task can move between Slack, VS Code/JetBrains, cloud, and terminal.

Apache-2.0 platform for stateful agents descended from MemGPT, distinguished by tiered self-editing memory persisted in a database (core memory in-context, archival out-of-context) plus 2026 research on git-based 'Context Repositories' and memory-native RL; in 2026 the company is consolidating around Letta Code, an open, model-agnostic memory-first coding agent harness (LettaBot archived May 2026 and folded in as Channels).

The NeurIPS-2024 research harness from the SWE-bench team that pioneered the agent-computer-interface concept: takes a GitHub issue and autonomously fixes it in a containerized environment with any LM, and remains the reference batch-evaluation harness for academic agent research.

MIT-licensed, type-safe Python agent framework from the Pydantic team with a FastAPI-like developer experience: schema-validated structured outputs, first-class MCP support, built-in OpenTelemetry observability (pairs with Logfire), and durable execution via native Temporal/DBOS integrations that offload model requests, tool calls, and MCP communication to fault-tolerant activities.

Terminal agent purpose-built for very large multi-file tasks — 2M-token effective context via tree-sitter project maps, a cumulative diff-review sandbox that stages changes outside the repo, and configurable full-auto mode — but Plandex Cloud wound down in Oct 2025 and the repo has had no pushes since, leaving it self-host-only and dormant.

MIT-licensed open agent harness from HKU's Data Intelligence Lab (agent loop, tool registry, permission system, hooks, skills, plugins, MCP, React TUI, automatic context compression) built for clarity and Claude-style workflow compatibility, shipping with 'Ohmo', a built-in personal agent that runs on existing Claude Code/Codex subscriptions and works via Feishu/Slack/Telegram/Discord, autonomously branching, coding, testing, and opening PRs.

The AutoGen + Semantic Kernel convergence (1.0 GA April 2026) for .NET and Python, whose BUILD 2026 'Agent Harness' made production patterns first-class — shell/filesystem access, human-in-the-loop approvals, context compaction, todo tracking, plan/execute modes, AgentSkillsProvider skill discovery, and BackgroundAgentsProvider subagents — with hosted execution via Microsoft Foundry.

GitHub's terminal agent (GA February 2026) that brings the Copilot coding agent to the CLI, distinguished by deep GitHub.com integration (built-in GitHub MCP server, in-terminal Issues/PRs/Gists tabs), an Autopilot autonomous mode, custom .agent.md agents, multi-vendor model choice, and AI-credit-based billing.

Anthropic's production agent library exposing the same harness that powers Claude Code (tool loop, file/bash/web tools, subagents, hooks, skills, MCP, permission checkpoints) in Python and TypeScript; the MIT-licensed SDK wraps the bundled Claude Code runtime and, as of 2026, underpins Anthropic Managed Agents with separate metered billing from June 15, 2026.

A deliberately radical ~100-line Python agent with bash as its only tool (no tool-calling API, no config sprawl) that still scores >74% on SWE-bench Verified, used by Meta, NVIDIA, and IBM as a baseline/benchmarking harness and proof that scaffolding can be minimal.

Mistral's minimal Apache-2.0 CLI coding agent (launched December 2025 alongside the open-weight Devstral 2 models, which score a vendor-reported 72.2% on SWE-bench Verified), distinguished by a deliberately lean self-hostable design — file read/write/patch, stateful shell, grep, and todo tools — with fully open, locally runnable model weights.

AWS's Claude-powered agentic IDE (VS Code fork; launched internationally May 7, 2026 as the successor to Amazon Q Developer) whose signature is spec-driven development — generating requirements.md/design.md/tasks.md with EARS acceptance criteria before coding — plus event-driven agent hooks, steering files, AGENTS.md/Skills.md/MCP support, and 'Kiro Powers' pre-packaged AWS MCP servers; the GitHub repo (3,980 stars) is a public issue tracker, not the source.

Moonshot AI's MIT-licensed TypeScript terminal agent (launched June 2026, replacing its earlier CLI) built around the token-efficient Kimi K2.7-Code 1T-parameter MoE model with 256K context, distinguished by built-in coder/explore/plan subagents running in isolated contexts and configurable OpenAI-compatible providers including local deployments.

Augment Code's terminal agent distinguished by its real-time Context Engine (semantic indexing of 1M+ files across repos, dependencies, docs, and git history), Unix-pipe-friendly print/quiet headless modes for CI/Lambda, custom slash commands, and credit-based plans from $20 (Indie) to enterprise.

A research-grade 'harness foundry' (MIT, v0.1.0 beta, arXiv paper 2606.14249) that composes agent harnesses from typed reusable processors across seven categories (context, control, evaluation, memory, multi-model, observability, tools) and evolves them via the AEGIS trace-driven engine, reporting an average +14.5% gain across ALFWorld, GAIA, WebShop, tau^3-Bench, and SWE-bench Verified without touching model weights.

Sourcegraph's opinionated no-model-picker coding agent (CLI + VS Code extension), distinguished by curated multi-model design with specialized subagents (Oracle for reasoning, Librarian for library research), Toolboxes (UNIX-script custom tools), skills with bundled MCP servers, and an ad-supported free mode alongside zero-markup pay-as-you-go credits.

Cursor's terminal/headless version of its IDE agent (shipped January 2026), distinguished by cloud handoff to isolated cloud VMs, sandboxed command execution with granular network/filesystem controls, hooks, async subagents, and shared MCP/rules/skills config with the Cursor editor.

Factory AI's agent-native development platform whose Droid CLI held top Terminal-Bench rankings (77.3% with GPT-5.3-Codex, Feb 2026), distinguished by deploying agents across CLI, IDEs, Slack, and Linear with broad multi-vendor model support including Chinese open-weight models; raised $150M at $1.5B in April 2026.

xAI's terminal coding agent (beta, launched June 5, 2026) distinguished by 8 parallel worktree-isolated subagents, a plan-first approval workflow, Arena Mode for competing outputs, an ACP-to-MCP bridge for MCP servers, OpenCode integration, and aggressive pricing via SuperGrok ($30/mo) and X Premium+ subscriptions.

VS Code-fork AI IDE whose Agent mode plus in-house Composer models, Background/Cloud Agents (isolated cloud VMs that report back as PRs), hooks, skills, and an agent SDK make it the most fully-built-out commercial agentic IDE; Composer 2 scored 61.7% on Terminal-Bench 2.0 and Composer 2.5 79.8% on SWE-bench Multilingual (SWE-bench Verified not reported).

ByteDance's VS Code-based AI IDE, notable for aggressive pricing (free tier plus $3-$100/mo token-based tiers, restructured Feb 2026) and the SOLO mode (standalone app since March 31, 2026) that acts as an end-to-end 'context engineer' handling requirements, code, terminal, browser testing and deployment, with MCP (stdio/SSE/streamable HTTP) and custom agents — but persistent privacy/telemetry concerns.

Google's agent-first development platform (launched Nov 2025, Antigravity 2.0 at I/O 2026 moved it toward a production standalone desktop app), distinguished by its Agent Manager 'mission control' that orchestrates parallel agents across editor, terminal, and an embedded browser subagent, with agents producing reviewable Artifacts (plans, screenshots, browser recordings) and a persistent knowledge/learnings base; runs Gemini 3 alongside Claude and GPT-OSS models.

OpenAI's delegation-first coding agent whose cloud mode runs each task in an isolated no-internet-by-default sandbox with the repo cloned, opens PRs, does automatic PR code review on GitHub and takes tasks from ChatGPT/Slack, bundled into ChatGPT Plus/Pro/Business with rolling credit-based billing since April 2026 (the CLI companion is open source, the cloud service is not).

YC-backed autonomous software engineer that picks up tickets from GitHub/Jira/Linear and opens tested PRs end-to-end across CLI, desktop, VS Code and cloud, distinguished by its own in-house Lumen model family (including on-device Scout and air-gapped sovereign variants for regulated industries like HSBC/NatWest/BAE) and the April 2026 'Swarm' feature for parallel long-horizon agents; credit-based seats at $20 (Hobby) and $200 (Professional).

Enterprise-focused 'Droids' platform ($1.5B valuation after April 2026 Series C): autonomous agents that execute refactors, migrations and incident investigations end-to-end across CLI, desktop, SDK and remote 'Droid Computers' cloud workloads, with both local and cloud background agents on the $20 Pro tier and adoption at Nvidia, Adobe, Morgan Stanley and MongoDB; notably model-agnostic across frontier providers.

Anthropic's non-technical desktop agent built on the Claude Code harness (no terminal required) that works over local folders, browsers (Claude in Chrome), and full computer use, with MCP connectors, Agent Skills, and plugins — expanded to web and mobile in July 2026 with cloud-run sessions for Max subscribers.