# AI Choice Engine coding harness catalog

> Where coding agents run, how they bill, and what to watch for. HTML: https://aichoiceengine.com/ai-harnesses

Cursor, Claude Code, GitHub Copilot, OpenAI Codex, Windsurf, and similar tools are compared here as products — not as the models they call.

## Claude Code

- Profile: https://aichoiceengine.com/ai-harnesses/claude-code
- Setup: https://aichoiceengine.com/ai-harnesses/setup/claude-code
- Vendor: Anthropic
- Surface: Terminal
- Pricing model: hybrid
- Autonomy: agentic
- Best for: Multi-file refactors; Migrations; Delegated engineering tasks
- Watch out: From 2026-08-14, new Pro/Max/Team sessions default to Auto (classifier-gated; classifier overhead free on those plans — Enterprise/API still opt-in). +50% weekly usage boost ends 2026-08-19 23:59 PT — don't size Pro/Max recommendations on temporary headroom. Included usage still shares allowance with Claude chat; an ANTHROPIC_API_KEY can bill API credits instead. Free Claude.ai does not include Claude Code. BAA + Claude Code requires ZDR, which blocks Covered Models (e.g. Fable 5) — plan HIPAA stacks accordingly.
- Verified: 2026-08-14

## Cursor

- Profile: https://aichoiceengine.com/ai-harnesses/cursor
- Setup: https://aichoiceengine.com/ai-harnesses/setup/cursor
- Vendor: Cursor
- Surface: Own IDE (VS Code fork)
- Pricing model: hybrid
- Autonomy: hybrid
- Best for: Daily feature work; Multi-file edits in-editor; Teams already on VS Code
- Watch out: Only works in its own IDE — adopting it means the whole team changes editor. Hobby is free but Agent/Tab are limited; Pro/Pro+/Ultra include usage pools with on-demand overage — the seat price is not an unlimited agent budget. Teams Standard ~$40 vs Premium ~$120 (dual usage pools). Composer Fast can burn ~6× the token rate of standard Composer — pair carefully with tokens-only Usage UI. SpaceX ~$60B Anysphere deal remains reported/pending close (regulatory); secondary coverage says Cursor brand may phase for some new products while the coding IDE keeps shipping near-term — ask about roadmap, DPA, support entity, and brand continuity before long contracts. Self-serve Usage UI is tokens-only (Spending/invoices still show $). Grok 4.6 is in the first-party pool. Launch week from 2026-08-12 (no published cutoff; still running as of 2026-08-14): 2× included usage, plus a 50% Grok 4.6 on-demand token discount on Teams — do not size seats on those promos. Cloud Agent Builds are included with Cloud Agents at no extra cost.
- Verified: 2026-08-14

## GitHub Copilot

- Profile: https://aichoiceengine.com/ai-harnesses/github-copilot
- Setup: https://aichoiceengine.com/ai-harnesses/setup/github-copilot
- Vendor: GitHub
- Surface: Plugin plus IDE and cloud agents
- Pricing model: hybrid
- Autonomy: hybrid
- Best for: Mixed-editor teams; Enterprise rollout; Procurement-led purchases
- Watch out: Capabilities vary by editor, plan, and agent surface (Neovim is completion-only). Individual tiers include Pro+ and Max (Max is 10,000 base + 10,000 flex = 20,000 total credits). Business/Enterprise promo AI-credit pools are 3,000 / 7,000 through 2026-09-01, then 1,900 / 3,900 standard. Annual Pro plans are retired for new buyers. MAI-Code-1-Flash retires 2026-09-10. Gemini 3.7 Flash is available from 2026-08-13 at provider list rates and is not the default.
- Verified: 2026-08-14

## OpenAI Codex

- Profile: https://aichoiceengine.com/ai-harnesses/openai-codex
- Setup: https://aichoiceengine.com/ai-harnesses/setup/openai-codex
- Vendor: OpenAI
- Surface: Terminal and cloud
- Pricing model: hybrid
- Autonomy: agentic
- Best for: Delegated coding tasks; Existing OpenAI commitments
- Watch out: ChatGPT plans include Codex with limits; API-key login is pure usage billing and does not unlock cloud features. One task can burn many credits. GPT-5.4 and GPT-5.4 mini retire from ChatGPT-authenticated Codex on 2026-08-31 (move those seats to GPT-5.6 Terra / Luna) — API-key Codex and the OpenAI Platform API are unaffected. Computer History on ChatGPT desktop (macOS) is opt-in and admin-gated; excluded from EEA/CH/UK at the 2026-08-13 launch.
- Verified: 2026-08-14

## Devin Desktop (formerly Windsurf)

- Profile: https://aichoiceengine.com/ai-harnesses/windsurf
- Setup: https://aichoiceengine.com/ai-harnesses/setup/windsurf
- Vendor: Cognition
- Surface: Own IDE
- Pricing model: hybrid
- Autonomy: hybrid
- Best for: Teams evaluating the current Devin Desktop surface; Teams migrating from Windsurf
- Watch out: Not Devin Cloud. Cascade reached EOL through July 2026 — Desktop’s current local agent is Devin Local. Quota can be shared with cloud sessions; confirm which surface you are buying.
- Verified: 2026-08-08

## Cline

- Profile: https://aichoiceengine.com/ai-harnesses/cline
- Setup: https://aichoiceengine.com/ai-harnesses/setup/cline
- Vendor: Cline
- Surface: Editor extension + CLI
- Pricing model: hybrid
- Autonomy: agentic
- Best for: VS Code users who want an open-source agent; Bring-your-own-model or ClinePass workflows; Teams avoiding a proprietary IDE lock-in
- Watch out: Agent loops burn tokens or Cline credits quickly; pick Provider / ClinePass / BYOK before sending private repos. ClinePass remains ~$9.99/mo with a $4.99 first-month promo — confirm live. Feature depth varies by surface.
- Verified: 2026-08-09

## Aider

- Profile: https://aichoiceengine.com/ai-harnesses/aider
- Setup: https://aichoiceengine.com/ai-harnesses/setup/aider
- Vendor: Aider
- Surface: Terminal
- Pricing model: usage-based
- Autonomy: hybrid
- Best for: Terminal-first developers; Open-source tooling; Git-native edit and review workflows
- Watch out: Applies edits and auto-commits by default — use --no-auto-commits for review-first sessions. You bring API keys; the repo map also sends structure to the model provider.
- Verified: 2026-08-08

## Devin Cloud

- Profile: https://aichoiceengine.com/ai-harnesses/devin-cloud
- Setup: https://aichoiceengine.com/ai-harnesses/setup/devin-cloud
- Vendor: Cognition
- Surface: Cloud agent
- Pricing model: hybrid
- Autonomy: agentic
- Best for: Delegated engineering work in a cloud workspace; Teams evaluating Cognition's agent surface
- Watch out: Not the same product as Devin Desktop; self-serve Pro/Max/Teams quota can be shared with Desktop. Long cloud sessions and retries burn the same allowance — confirm which Cognition surface you are buying before rollout.
- Verified: 2026-08-08

## Replit

- Profile: https://aichoiceengine.com/ai-harnesses/replit
- Setup: https://aichoiceengine.com/ai-harnesses/setup/replit
- Vendor: Replit
- Surface: Browser
- Pricing model: hybrid
- Autonomy: hybrid
- Best for: Prototyping; Teaching; Contributors who cannot install a toolchain
- Watch out: Starter is Lite-build only (Plan mode needs Core). GitHub import works; offline-first and heavy local-only workflows still fit poorly. Cloud autoscale/storage/transfer rates were cut 2026-08-01 — re-check hosting math.
- Verified: 2026-08-08

## Lovable

- Profile: https://aichoiceengine.com/ai-harnesses/lovable
- Setup: https://aichoiceengine.com/ai-harnesses/setup/lovable
- Vendor: Lovable
- Surface: Browser app builder
- Pricing model: usage-based
- Autonomy: agentic
- Best for: Fast web prototypes; Internal tools; Teams that want generated code with browser previews
- Watch out: From 2026-09-09, Free/Pro may train unless you opt out. Also review generated code, backend boundaries, Row Level Security, GitHub permissions, and deployment costs before production use.
- Verified: 2026-08-08

## Bolt

- Profile: https://aichoiceengine.com/ai-harnesses/bolt
- Setup: https://aichoiceengine.com/ai-harnesses/setup/bolt
- Vendor: StackBlitz
- Surface: Browser app builder
- Pricing model: usage-based
- Autonomy: agentic
- Best for: In-browser prototypes; StackBlitz-native iteration; Early product exploration
- Watch out: Treat generated auth, env, and deploy settings as a first draft. Token/credit burn on retries is the real cost. ACE has no publisher affiliate programme for Bolt.
- Verified: 2026-08-13

## Muse Code

- Profile: https://aichoiceengine.com/ai-harnesses/muse-code
- Setup: https://aichoiceengine.com/ai-harnesses/setup/muse-code
- Vendor: Meta
- Surface: Terminal
- Pricing model: usage-based
- Autonomy: agentic
- Best for: Large-repo agentic coding; Long-running tool loops; Teams willing to bill Meta Model API usage
- Watch out: macOS/Linux beta only; Meta account required. Co-trained for Muse Spark — swapping models is not the product story. Contributor pricing trains on your usage and carries lower RPM than standard API rates.
- Verified: 2026-08-06

## Grok Build

- Profile: https://aichoiceengine.com/ai-harnesses/grok-build
- Setup: https://aichoiceengine.com/ai-harnesses/setup/grok-build
- Vendor: SpaceXAI
- Surface: Terminal
- Pricing model: hybrid
- Autonomy: agentic
- Best for: Parallel isolated subagents; Teams already on SuperGrok; Plan-then-diff workflows
- Watch out: x.ai currently advertises Grok Build as available to try for Free; fuller parallel capacity may still sit on higher SuperGrok / X Premium+ tiers. Early product relative to Claude Code. Included usage is 2× for the first week from 2026-08-12 — do not size seats on that promo.
- Verified: 2026-08-14

## OpenCode

- Profile: https://aichoiceengine.com/ai-harnesses/opencode
- Setup: https://aichoiceengine.com/ai-harnesses/setup/opencode
- Vendor: OpenCode
- Surface: Terminal, desktop beta, and IDE extensions
- Pricing model: usage-based
- Autonomy: agentic
- Best for: Terminal-first open-source workflows; Multi-provider model flexibility; Teams that want Plan mode before edits
- Watch out: You pay whichever model provider you connect (or Zen/Go). OpenCode Go is not blanket ZDR — Grok 4.5 / GPT-5.6 Luna retain ~30 days; DeepSeek V4 Flash ZDR only through 2026-08-31. Claude Pro/Max OAuth in third-party harnesses violates Anthropic ToS — use API keys. Windows beginners often prefer WSL2.
- Verified: 2026-08-09

## Kilo Code

- Profile: https://aichoiceengine.com/ai-harnesses/kilocode
- Setup: https://aichoiceengine.com/ai-harnesses/setup/kilocode
- Vendor: Kilo
- Surface: VS Code / JetBrains extension + CLI
- Pricing model: hybrid
- Autonomy: agentic
- Best for: IDE agent workflows without leaving VS Code or JetBrains; Teams migrating from Roo Code; CLI scripting and bring-your-own-provider setups
- Watch out: Not terminal-only: Marketplace IDE is the flagship for many buyers. Billing spans free platform, Gateway/Pass, BYOK, and optional cloud compute.
- Verified: 2026-08-08

## Continue

- Profile: https://aichoiceengine.com/ai-harnesses/continue
- Setup: https://aichoiceengine.com/ai-harnesses/setup/continue
- Vendor: Continue
- Surface: VS Code and JetBrains extension (final OSS release)
- Pricing model: usage-based
- Autonomy: hybrid
- Best for: Teams that already run Continue locally; Local-model and BYO-key workflows on a frozen OSS release; Migration comparisons toward Cline, OpenCode, Kilo, or Cursor
- Watch out: Acquired by Cursor (Jun 2026): repo read-only, Hub/cloud wound down after Jul 15. Prefer migration alternatives for new team rollouts; JetBrains maintainers recommend the Continue CLI (cn) over the plugin.
- Verified: 2026-08-12

## Amp

- Profile: https://aichoiceengine.com/ai-harnesses/amp
- Setup: https://aichoiceengine.com/ai-harnesses/setup/amp
- Vendor: Amp
- Surface: Terminal + IDE integrations
- Pricing model: hybrid
- Autonomy: agentic
- Best for: Terminal-first agent work with IDE hooks; Teams evaluating Amp workflows alongside other terminal agents; Remote-controlled agent sessions via the web
- Watch out: Prefer the official Amp install script over unrelated npm packages named amp. Subscriptions are Megawatt $20 and Gigawatt $200. Orbs a1 sizes bill about $0.08–$1.32/hr, auto-pause after 5 minutes idle, and can be set per thread with `--orb-size`. Attach Anything, subscription Orbs, and linked ChatGPT usage are separate from the free sign-in path — confirm live Amp billing before a team pilot. A linked ChatGPT subscription bills Dial low/medium/high to that sub (Terra/Sol); ultra still uses Fable on Amp credits.
- Verified: 2026-08-14

## Goose

- Profile: https://aichoiceengine.com/ai-harnesses/goose
- Setup: https://aichoiceengine.com/ai-harnesses/setup/goose
- Vendor: Block
- Surface: Terminal + desktop
- Pricing model: usage-based
- Autonomy: agentic
- Best for: Open-source agent pilots; Multi-provider / local-model experiments; Teams that want CLI + optional desktop
- Watch out: You bring inference (cloud or local). Desktop and CLI install paths differ; confirm PATH after install and run goose configure before the first agent session.
- Verified: 2026-08-08

## Warp Agent CLI

- Profile: https://aichoiceengine.com/ai-harnesses/warp
- Setup: https://aichoiceengine.com/ai-harnesses/setup/warp
- Vendor: Warp
- Surface: Terminal agent CLI (any emulator)
- Pricing model: hybrid
- Autonomy: agentic
- Best for: Developers who want Warp Agent without switching terminal apps; SSH and remote terminal agent sessions; Teams already on Warp plans
- Watch out: Warp account/plan limits apply to the CLI. Zero Data Retention on Warp-managed inference is advertised for Business and Enterprise plans, not all tiers — still confirm BYOK and cloud-agent paths separately. Do not confuse the Agent CLI installer with the full Warp terminal app cask — they are related products with different install paths. Claude Fable 5 and Opus 5 are 50% off for a limited time with no published end date — confirm the live Warp billing page before a pilot.
- Verified: 2026-08-14

## DeepSeek Harness

- Profile: https://aichoiceengine.com/ai-harnesses/deepseek-harness
- Setup: https://aichoiceengine.com/ai-harnesses/setup/deepseek-harness
- Vendor: DeepSeek
- Surface: Local web UI (CLI launcher)
- Pricing model: usage-based
- Autonomy: agentic
- Best for: Teams evaluating an open DeepSeek-native agent loop; Plugin authors targeting the dsh-plugin ecosystem; Local web-UI agent sessions with bring-your-own keys
- Watch out: v0.1 developer preview (`@deepseek-ai/dsh` 0.1.0-rc.6) with compatibility-breaking changes. Requires Node.js ^22.19 or >=24. The client is free; inference is billed by the provider key you paste in Settings → Models (stored under $DSH_HOME/.credentials.yaml). Do not install the PyPI package named deepseek-harness — the official launcher is `npx @deepseek-ai/dsh web`. Do not treat Google- or vendor-blog agent scores as a DeepSeek Harness product benchmark.
- Verified: 2026-08-14