A curated, verified, cross-vendor directory of AI agent harnesses — the runtime scaffolding that turns a model into an agent.
◆ 74 harnesses profiled · 9-axis capability schema · last verified 2026-07-08
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.
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.
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.
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.
The Cline/Roo merger-fork turned all-in-one agentic platform (~26k stars, 3M+ users): IDE extension, CLI, and hosted KiloClaw background agent reachable via Telegram/Discord/Slack, with 500+ models, mid-task model switching, an MCP marketplace, subagents (GA April 2026), and optional Kilo Pass credits from $19/mo.
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).
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).
The Windsurf AI IDE's Cascade agent, post-Cognition-acquisition, was rebranded to Devin Desktop (June 2, 2026) with Cascade deprecated (available through July 1, 2026) in favor of the Rust-rewritten 'Devin Local' agent (30% more token-efficient, subagents), pairing local agentic editing (Fast Context codebase awareness, MCP, Memories, Workflows) with Devin cloud sessions for background work.
Copilot's agent mode inside VS Code (extension itself MIT open-sourced; backend service proprietary) iterates on errors, runs tools/terminal, and applies multi-file edits, distinguished by deep GitHub-ecosystem reach: MCP support, SKILL.md Agent Skills (added April 2026), custom .agent.md agents, a multi-vendor model picker with Ollama BYOK, and a cloud 'coding agent' that works issues into PRs on GitHub Actions.
JetBrains' autonomous coding agent for IntelliJ-family IDEs (plus a Junie CLI that left beta March 2026 with BYOK/OpenRouter routing and CI/CD support), distinguished by leverage of the IDE's code index, a confirmation-gated loop with opt-in 'Brave Mode', Docker/worktree-isolated parallel tasks, remote cloud execution mode, and a companion Claude Agent built on Anthropic's Agent SDK.
Google's asynchronous coding agent that clones your GitHub repo into a secure cloud VM, plans and executes tasks (bug fixes, tests) and returns PRs; free tier of 15 tasks/day with higher quotas bundled into Google AI Pro/Ultra, Gemini 3.1 Pro/3 Flash under the hood, curated MCP connections (Linear, Supabase, Neon, Context7) added Feb 2026 and an alpha REST API for Slack/Linear-driven automation.
The viral general-purpose autonomous agent that plans and executes multi-step tasks in a cloud VM (browsing, code, docs, slides), notable for Wide Research parallel sub-agents and for Meta's $2B acquisition being blocked by China's NDRC in April 2026.
Credit-based all-in-one 'AI employee' workspace whose Super Agent mixes multiple frontier models with 80+ tools to produce slides, sheets, sites, and calls, and whose 2026 'Genspark Claw' gives each user a persistent cloud computer commandable from WhatsApp/Telegram/Slack/Teams.
Cloud web-agent from DeepMind-alumni startup H Company ($220M seed) that orchestrates sub-agents over its own Holo vision-language action models (92.2% WebVoyager at ~$0.13/task via open-sourced Surfer-H), with Holo 3 in 2026 targeting any-UI enterprise automation.