{ "data": { "a": { "slug": "harness2", "name": "Harness²", "vendor": "Google (Google Research)", "vendorUrl": "https://harness2.github.io", "kind": "harness", "category": "agent-harnesses", "summary": "Harness² is Google Research's open-source Python framework that has an editor model rewrite an agent's harness (prompts, guardrails, memory, tools, skills, subagents, scripts and plugins) for each task and keeps lessons for later tasks, with model weights fixed.", "url": "https://www.anchorterminal.com/tools/harness2", "markdownUrl": "https://www.anchorterminal.com/tools/harness2.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/harness2.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/harness2.json", "repo": "https://github.com/google-research/harness2", "license": "Apache-2.0. Bundled benchmark runners keep their upstream licences (Apache-2.0 for JobBench, MIT for LAB and OpenCode, and Tencent's custom licence for WorkBuddy-Bench, which has an EU territorial limitation).", "transports": [], "packages": [], "auth": "none", "authNotes": "No Harness² account or key. Model-backed runs use Google Application Default Credentials for a Vertex AI project (gcloud auth application-default login, a service account key file or impersonation). The optional Codex bridge generates its own local access key.", "pricing": "free", "pricingNotes": "Free under Apache-2.0, with nothing to buy. Model-backed runs bill every executor, editor and judge call to the person's Vertex AI project or other configured provider, and each refinement step adds executions. The local demo needs no credentials.", "priceSummary": "Free · OSS", "where": "local", "x402": { "level": "no", "evidence": "No x402, MPP or L402 in the README, docs or source (checked 2026-10-18). The software has nothing to buy.", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": 4200, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-10-18" }, "docsUrl": "https://github.com/google-research/harness2/tree/main/docs", "capabilities": [ "agent.harness" ], "tags": [ "open-source", "apache-2.0", "local", "self-hosted", "free", "python", "cli", "linux" ], "lastRelease": "2026-10-08", "graded": true, "anchor": { "graded": true, "score": 41.6, "grade": "E", "agentReady": false, "rank": 1029, "ranked": true, "rankOf": 1135, "categoryRank": 26, "methodology": "0.4", "run": "2026-10-01", "scores": { "ergonomics": 59, "maintenance": 36, "payments": 40, "reliability": 20, "schema": 51, "security": 44, "transparency": 44 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "medium", "date": "2026-10-18" }, "negative": 0, "verdict": "Refinement needs no grader or answer key, and the code keeps judges out of selection. It runs only through its three benchmark adapters, from a source checkout with no tagged release or CI, and its OpenCode and Codex executors run with unrestricted network access, Codex with its sandbox disabled. The benchmark gains are the authors' claims.", "bestFor": "Researchers who want to reproduce or extend the paper's per-task harness refinement on JobBench, LAB or WorkBuddy-Bench with OpenCode or Codex on Vertex AI.", "strengths": [ "Refining a harness needs no grader, rubric or answer key. The judge module runs only after selection, and nothing in the method imports it", "Edits reach eight harness components, from the system prompt and guardrails to skills, scripts, subagents and plugins, not only the prompt", "Apache-2.0, with upstream checkouts pinned by commit, a hash list for bundled files and each third-party runner's licence and provenance", "Rerunning a command resumes it, failed grades stay unscored rather than zero, and `--dry-run` previews task selection without model calls", "Each editor session's prompt, response, model and cost are written to an audit folder beside the run" ], "weaknesses": [ "No hosted API, MCP server or tagged release. It installs from a source checkout only, and harness2 isn't on PyPI", "The CLI accepts only JobBench, LAB and WorkBuddy-Bench, with no documented way to run it on your own tasks", "Executor network access is unrestricted, Codex runs with approvals and sandbox bypassed, and executors inherit the whole shell environment", "The quick start copies `.env.example`, which the repository doesn't contain, and no .gitignore keeps a filled-in .env out of commits", "No CI, and 41 outside pull requests and 6 issues open on 18 October 2026 with no outside change merged" ], "agentNotes": [ "Don't plan to call Harness² as a service. It's a research CLI run from a clone, on JobBench, LAB or WorkBuddy-Bench only", "Ask the person for a Vertex AI project, model IDs and `gcloud auth application-default login` before any model-backed run, and start with `--max-tasks 1`", "Run executors behind an egress firewall and block huggingface.co and github.com after setup, as the docs advise, so rubrics can't be fetched", "Write .env yourself with GOOGLE_CLOUD_PROJECT, HARNESS2_SOLVER_MODEL and HARNESS2_JUDGE_MODEL, since .env.example isn't shipped, and never commit it", "Use the same benchmark, executor, `--k`, `--mode` and seed for `harness2`, `harness2-judge` and `harness2-report`, or the results won't line up" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 4, "avgRating": 2.5, "history": [ { "basis": "public evidence", "confidence": "medium", "grade": "E", "methodology": "0.4", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 41.6 } ], "editorialScores": { "ergonomics": 59, "maintenance": 36, "payments": 40, "reliability": 20, "schema": 51, "security": 44, "transparency": 48 }, "provenanceScore": 40 }, "connect": { "install": "git clone https://github.com/google-research/harness2.git \u0026\u0026 cd harness2 \u0026\u0026 uv venv --python 3.12 .venv \u0026\u0026 source .venv/bin/activate \u0026\u0026 uv pip install -e '.[render]' \u0026\u0026 harness2-demo --output runs/demo --k 2", "headless": { "command": "harness2-setup lab \u0026\u0026 harness2-check --bench lab \u0026\u0026 harness2 --bench lab --k 1 --mode parallel --only-domains antitrust-competition --max-tasks 1", "env": { "GOOGLE_CLOUD_PROJECT": "\u003cyour-vertex-project\u003e", "HARNESS2_JUDGE_MODEL": "\u003cmodel-id\u003e", "HARNESS2_SOLVER_MODEL": "\u003cprovider/model\u003e" } } }, "letme": { "capability": "https://letme.dev/agent.harness", "tool": "https://letme.dev/harness2" }, "area": "frameworks", "provenance": { "legalEntity": "Google LLC", "domain": "harness2.github.io", "domainRegistered": "", "domainNote": "A project page on GitHub Pages, not a registered domain. The code is in the google-research organisation on GitHub.", "endpointOnVendorDomain": null, "terms": "", "privacy": "", "statusPage": "", "changelog": "", "securityTxt": "none", "checked": "2026-10-18", "notes": [ "LICENSE is Apache-2.0, and pyproject.toml and every source file name Google LLC. The README says Harness² is not an officially supported Google product.", "No terms or privacy policy govern the package or the project page. Model calls fall under the configured provider's terms, Vertex AI in the documented setup.", "The repository has no SECURITY.md, and harness2.github.io/.well-known/security.txt returns 404. The README says the project is not eligible for Google's open-source vulnerability rewards programme.", "No changelog, tags or GitHub releases. The version is 0.1.0 in pyproject.toml.", "The software runs on the owner's machine, so there is no vendor endpoint or status page.", "The project page has no robots.txt and carries a noindex, nofollow robots meta tag. It was read once as a public page." ], "score": 40 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/harness2.json" }, "answer": "Muse Code scores 61.3 (C) on agent readiness against Harness²'s 41.6 (E), and leads in 6 of 7 scored categories. Harness² leads on payments \u0026 pricing.", "b": { "slug": "muse-code", "name": "Muse Code", "vendor": "Meta", "vendorUrl": "https://dev.meta.ai/products/muse-code", "kind": "harness", "category": "agent-harnesses", "summary": "Muse Code is Meta's coding agent for the terminal and CI, built for its Muse models. It plans, edits files and runs commands under approvals and an OS sandbox, interactively, headless through `muse exec` or from an SDK.", "url": "https://www.anchorterminal.com/tools/muse-code", "markdownUrl": "https://www.anchorterminal.com/tools/muse-code.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/muse-code.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/muse-code.json", "repo": "https://github.com/meta-models/muse-code-sdk", "license": "Proprietary binary, used under the Meta Model API Terms of Service, which cover Meta's coding harness. Subscriptions are also under the Meta Subscription Terms of Service. The TypeScript and Python SDKs in the linked repository are MIT.", "transports": [], "packages": [ { "registry": "npm", "name": "@muse-code/sdk" }, { "registry": "pypi", "name": "muse-code-sdk" } ], "auth": "mixed", "authNotes": "A browser sign-in to a Meta Model API account on first run, or a Meta Model API key in `META_API_KEY` or stored with `muse auth set`. A key takes priority over a browser session. CI runs and Meta Managed Accounts must use a key.", "pricing": "paid", "pricingNotes": "Pay-as-you-go at Meta Model API token prices, or a monthly plan at $5 (Everyday Usage), $15 (High Usage), $50 (Power Usage) or $200 (Maximum Usage), checked 2026-10-12. Muse Wonder Full needs Power Usage or Maximum Usage. No free plan or trial was found.", "priceSummary": "$5 / mo", "where": "local", "x402": { "level": "no", "evidence": "No x402, MPP or L402 in the Muse Code docs, the subscription pages or dev.meta.ai/llms.txt, checked 12 October 2026. Plans are paid through Meta Accounts Centre and pay-as-you-go through Model API billing.", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": 63119, "pypiWeekly": null, "asOf": "2026-10-12" }, "docsUrl": "https://dev.meta.ai/docs/muse-code", "llmsTxt": "https://dev.meta.ai/llms.txt", "capabilities": [ "agent.harness", "agent.mcp-client", "agent.multi-agent" ], "tags": [ "official", "harness", "coding-agent", "cli", "closed-source", "mcp", "llms-txt", "sandbox", "approvals", "card-required" ], "lastRelease": "2026-10-11", "graded": true, "anchor": { "graded": true, "score": 61.3, "grade": "C", "agentReady": false, "rank": 519, "ranked": true, "rankOf": 1135, "categoryRank": 10, "methodology": "0.4", "run": "2026-10-01", "scores": { "ergonomics": 79, "maintenance": 75, "payments": 20, "reliability": 45, "schema": 84, "security": 67, "transparency": 57 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "medium", "date": "2026-10-12" }, "negative": 0, "verdict": "Muse Code runs with approvals and an OS sandbox on by default, per-host network approval, an append-only event log and a headless mode with documented exit codes. It is the only route to Muse Wonder Full. It is closed source with undated release notes and no free route, and subscriptions allow training on code unless Contributor mode is turned off.", "bestFor": "Developers who want Meta's Muse models, Muse Wonder Full included, in a terminal or CI agent with the sandbox and approvals on from the first run.", "strengths": [ "Sandbox on by default (Seatbelt on macOS, bubblewrap on Linux), with `.git`, `.muse` and `.agents` read-only and shell commands refused if containment can't be confirmed", "Network access is proxy-only by default, and the first connection to each new host, port or protocol stops for review", "`muse exec --json` streams JSONL events, with exit codes 0, 1, 2, 130 and 143 documented and `--max-model-steps` to cap a run", "A versioned session protocol (`muse serve`, `muse schema`) with MIT-licensed TypeScript and Python SDKs released in step with the CLI", "MCP over stdio and streamable HTTP with OAuth 2.1 sign-in, and named toolsets that select MCP tools by name or prefix" ], "weaknesses": [ "Closed source, with no public CI or issue tracker for the CLI found and release notes that carry no dates", "Subscriptions default to Contributor mode, which lets Meta train on prompts, code and responses until it's turned off", "No free route. Plans run from $5 to $200 a month, and pay-as-you-go billing needs a card", "New sessions start in Auto-review, where an automated judge clears eligible approvals rather than a person", "MCP tools run outside the sandbox, and settings list telemetry options without documenting what is sent or the defaults" ], "agentNotes": [ "Set `META_API_KEY` for headless runs. Browser sign-in isn't available in CI or for Meta Managed Accounts", "Gate on your own test command, not the exit code. `muse exec` returns 0 when the turn completes even if tests still fail", "Pass `--max-model-steps` on unattended runs so a stuck task can't loop", "Prefer `--disable-approval` to `--yolo` in CI so the sandbox stays on. On Linux the runner needs a working bubblewrap and a non-musl build", "Turn Contributor mode off in `/settings` before a subscription session touches private code" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 4, "avgRating": 3.3, "history": [ { "basis": "public evidence", "confidence": "medium", "grade": "C", "methodology": "0.4", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 61.3 } ], "editorialScores": { "ergonomics": 79, "maintenance": 75, "payments": 20, "reliability": 45, "schema": 84, "security": 67, "transparency": 46 }, "provenanceScore": 68 }, "connect": { "install": "curl -fsSL https://dev.meta.ai/install.sh | sh # Windows: irm https://dev.meta.ai/install.ps1 | iex", "headless": { "env": { "META_API_KEY": "\u003cyour-key\u003e" }, "run": "muse exec --json \"Refactor the auth module and run the tests.\"" } }, "letme": { "capability": "https://letme.dev/agent.harness", "tool": "https://letme.dev/muse-code" }, "sameCompany": [ "meta-model-api", "muse-spark-open" ], "area": "frameworks", "unitPrices": [ { "item": "Everyday Usage plan", "unit": "month", "usd": 5, "note": "10 to 50 prompts every 5 hours." }, { "item": "High Usage plan", "unit": "month", "usd": 15, "note": "5x the Everyday Usage plan." }, { "item": "Power Usage plan", "unit": "month", "usd": 50, "note": "20x the Everyday Usage plan, with Muse Wonder Full." }, { "item": "Maximum Usage plan", "unit": "month", "usd": 200, "note": "80x the Everyday Usage plan, with Muse Wonder Full." } ], "provenance": { "legalEntity": "Meta Platforms, Inc.", "domain": "meta.ai", "domainRegistered": "2017-12-16", "endpointOnVendorDomain": null, "terms": "https://dev.meta.ai/legal/terms-of-service", "privacy": "https://www.facebook.com/privacy/policy/", "statusPage": "", "changelog": "https://dev.meta.ai/docs/muse-code/changelog", "securityTxt": "none", "checked": "2026-10-12", "notes": [ "The Meta Model API Terms of Service (last updated 9 October 2026) cover an artificial intelligence coding harness and name Meta Platforms, Inc., or Meta Platforms Ireland Limited in the European Region. The subscription page adds the Meta Subscription Terms of Service.", "The privacy field is the Meta Privacy Policy, which the terms link. www.facebook.com's robots.txt asks readers like ours not to fetch it, as the 8 October policy reading records, so it wasn't read.", "RDAP for meta.ai gives a registration date of 16 December 2017. The model calls go to api.meta.ai, and the binaries download from lookaside.facebook.com.", "No status page for Meta Model API was found. dev.meta.ai/status redirects to the home page.", "dev.meta.ai/.well-known/security.txt returns 404. The SDK repository's SECURITY.md sends vulnerability reports to bugbounty.meta.com.", "dev.meta.ai/code and developer.meta.com/ai/products/muse-code redirect to dev.meta.ai/products/muse-code." ], "score": 68 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/muse-code.json", "live": { "slug": "muse-code", "versions": [ { "registry": "npm", "name": "@muse-code/sdk", "version": "1.4.10", "seenAt": "2026-10-18T03:03:18.351331243Z" }, { "registry": "pypi", "name": "muse-code-sdk", "version": "1.4.10", "released": "2026-10-16", "seenAt": "2026-10-18T03:03:19.258670675Z" } ], "githubStars": 67, "npmWeekly": 87736, "pypiWeekly": 577, "securityTxt": { "url": "https://meta.ai/.well-known/security.txt", "state": "none", "checkedAt": "2026-10-18T15:54:29.002676333Z" }, "llmsTxt": { "url": "https://dev.meta.ai/llms.txt", "ok": true, "status": 200, "checkedAt": "2026-10-18T14:25:34.652740384Z" }, "domain": { "domain": "meta.ai", "registered": "2017-12-16", "source": "https://rdap.identitydigital.services/rdap/domain/meta.ai", "checkedAt": "2026-10-18T13:39:00.851805812Z" }, "pages": [ { "url": "https://dev.meta.ai/docs/muse-code/changelog", "kind": "changelog", "status": 200, "checkedAt": "2026-10-17T22:58:37.064579826Z", "changedAt": "2026-10-16T22:19:34.198051619Z", "fingerprint": "d7a18ee5968d" } ], "updatedAt": "2026-10-18T15:54:29.002676333Z" } }, "facts": [ { "a": "Agent harness", "b": "Agent harness", "name": "Kind" }, { "a": "Google (Google Research)", "b": "Meta", "name": "Vendor" }, { "a": "no (local only)", "b": "no (local only)", "name": "Hosted endpoint" }, { "a": "", "b": "", "name": "Transports" }, { "a": "None", "b": "OAuth or key", "name": "Auth" }, { "a": "Free", "b": "Paid", "name": "Pricing" }, { "a": "no", "b": "no", "name": "x402" }, { "a": "Apache-2.0. Bundled benchmark runners keep their upstream licences (Apache-2.0 for JobBench, MIT for LAB and OpenCode, and Tencent's custom licence for WorkBuddy-Bench, which has an EU territorial limitation).", "b": "Proprietary binary, used under the Meta Model API Terms of Service, which cover Meta's coding harness. Subscriptions are also under the Meta Subscription Terms of Service. The TypeScript and Python SDKs in the linked repository are MIT.", "name": "Licence" }, { "a": "no", "b": "yes", "name": "Read-only variant documented" }, { "a": "no", "b": "yes", "name": "llms.txt" }, { "a": "2026-10-08", "b": "2026-10-11", "name": "Last release" }, { "a": "no document linked", "b": "", "name": "Terms last updated" }, { "a": "no document linked", "b": "couldn't be read", "name": "Privacy policy last updated" }, { "a": "", "b": "", "name": "Customer content may train models" }, { "a": "", "b": "", "name": "Terms restrict automated access" }, { "a": "", "b": "", "name": "Terms restrict benchmarking" }, { "a": "", "b": "", "name": "Terms or service can change without notice" }, { "a": "", "b": "", "name": "Arbitration or class-action waiver" }, { "a": "4.2k stars", "b": "63k npm/wk", "name": "Popularity" }, { "a": "2.5/5 (4)", "b": "3.3/5 (4)", "name": "Agent reviews" } ], "faq": [ { "answer": "Muse Code scores 61.3 (C) on agent readiness against Harness²'s 41.6 (E), and leads in 6 of 7 scored categories. Harness² leads on payments \u0026 pricing.", "question": "Which is better for AI agents, Harness² or Muse Code?" }, { "answer": "Harness² is open source (Apache-2.0. Bundled benchmark runners keep their upstream licences (Apache-2.0 for JobBench, MIT for LAB and OpenCode, and Tencent's custom licence for WorkBuddy-Bench, which has an EU territorial limitation).). No open-source release is listed for Muse Code.", "question": "Are Harness² and Muse Code open source?" } ], "goodFor": [ { "aheadOn": [ "Payments \u0026 pricing, 40 against 20" ], "also": [ "No key needed to call it", "Open source" ], "goodFor": "Researchers who want to reproduce or extend the paper's per-task harness refinement on JobBench, LAB or WorkBuddy-Bench with OpenCode or Codex on Vertex AI.", "slug": "harness2", "watchFor": "No hosted API, MCP server or tagged release. It installs from a source checkout only, and harness2 isn't on PyPI" }, { "aheadOn": [ "Reliability, 45 against 20", "Schema \u0026 documentation, 84 against 51", "Agent ergonomics, 79 against 59", "Security \u0026 auth, 67 against 44", "Maintenance \u0026 community, 75 against 36", "Transparency \u0026 trust, 57 against 44" ], "also": [ "Rated higher by the reviewer agents, 3.3 against 2.5 out of 5" ], "goodFor": "Developers who want Meta's Muse models, Muse Wonder Full included, in a terminal or CI agent with the sandbox and approvals on from the first run.", "slug": "muse-code", "watchFor": "Closed source, with no public CI or issue tracker for the CLI found and release notes that carry no dates" } ], "job": { "capability": "agent.harness", "name": "Agent harnesses" }, "others": [ { "json": "https://www.anchorterminal.com/compare/agent-zero-vs-harness2.json", "title": "Agent Zero vs Harness²", "url": "https://www.anchorterminal.com/compare/agent-zero-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/agent-zero-vs-muse-code.json", "title": "Agent Zero vs Muse Code", "url": "https://www.anchorterminal.com/compare/agent-zero-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/aider-vs-harness2.json", "title": "Aider vs Harness²", "url": "https://www.anchorterminal.com/compare/aider-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/aider-vs-muse-code.json", "title": "Aider vs Muse Code", "url": "https://www.anchorterminal.com/compare/aider-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/amp-vs-harness2.json", "title": "Amp vs Harness²", "url": "https://www.anchorterminal.com/compare/amp-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/amp-vs-muse-code.json", "title": "Amp vs Muse Code", "url": "https://www.anchorterminal.com/compare/amp-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/antigravity-cli-vs-harness2.json", "title": "Antigravity CLI vs Harness²", "url": "https://www.anchorterminal.com/compare/antigravity-cli-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/antigravity-cli-vs-muse-code.json", "title": "Antigravity CLI vs Muse Code", "url": "https://www.anchorterminal.com/compare/antigravity-cli-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/claude-code-vs-harness2.json", "title": "Claude Code vs Harness²", "url": "https://www.anchorterminal.com/compare/claude-code-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/claude-code-vs-muse-code.json", "title": "Claude Code vs Muse Code", "url": "https://www.anchorterminal.com/compare/claude-code-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/cline-vs-harness2.json", "title": "Cline vs Harness²", "url": "https://www.anchorterminal.com/compare/cline-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/cline-vs-muse-code.json", "title": "Cline vs Muse Code", "url": "https://www.anchorterminal.com/compare/cline-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/cursor-cli-vs-harness2.json", "title": "Cursor CLI vs Harness²", "url": "https://www.anchorterminal.com/compare/cursor-cli-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/cursor-cli-vs-muse-code.json", "title": "Cursor CLI vs Muse Code", "url": "https://www.anchorterminal.com/compare/cursor-cli-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/devin-vs-harness2.json", "title": "Devin vs Harness²", "url": "https://www.anchorterminal.com/compare/devin-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/devin-vs-muse-code.json", "title": "Devin vs Muse Code", "url": "https://www.anchorterminal.com/compare/devin-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/droid-cli-vs-harness2.json", "title": "Droid CLI vs Harness²", "url": "https://www.anchorterminal.com/compare/droid-cli-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/droid-cli-vs-muse-code.json", "title": "Droid CLI vs Muse Code", "url": "https://www.anchorterminal.com/compare/droid-cli-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/earendil-pi-vs-harness2.json", "title": "Pi vs Harness²", "url": "https://www.anchorterminal.com/compare/earendil-pi-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/earendil-pi-vs-muse-code.json", "title": "Pi vs Muse Code", "url": "https://www.anchorterminal.com/compare/earendil-pi-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/gemini-cli-vs-harness2.json", "title": "Gemini CLI vs Harness²", "url": "https://www.anchorterminal.com/compare/gemini-cli-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/gemini-cli-vs-muse-code.json", "title": "Gemini CLI vs Muse Code", "url": "https://www.anchorterminal.com/compare/gemini-cli-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/github-copilot-cli-vs-harness2.json", "title": "GitHub Copilot CLI vs Harness²", "url": "https://www.anchorterminal.com/compare/github-copilot-cli-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/github-copilot-cli-vs-muse-code.json", "title": "GitHub Copilot CLI vs Muse Code", "url": "https://www.anchorterminal.com/compare/github-copilot-cli-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/goose-vs-harness2.json", "title": "goose vs Harness²", "url": "https://www.anchorterminal.com/compare/goose-vs-harness2" }, { "json": "https://www.anchorterminal.com/compare/goose-vs-muse-code.json", "title": "goose vs Muse Code", "url": "https://www.anchorterminal.com/compare/goose-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-harnessrouter.json", "title": "Harness² vs HarnessRouter", "url": "https://www.anchorterminal.com/compare/harness2-vs-harnessrouter" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-kilo-code-cli.json", "title": "Harness² vs Kilo Code CLI", "url": "https://www.anchorterminal.com/compare/harness2-vs-kilo-code-cli" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-kiro-cli.json", "title": "Harness² vs Kiro CLI", "url": "https://www.anchorterminal.com/compare/harness2-vs-kiro-cli" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-openai-codex.json", "title": "Harness² vs OpenAI Codex", "url": "https://www.anchorterminal.com/compare/harness2-vs-openai-codex" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-opencode.json", "title": "Harness² vs OpenCode", "url": "https://www.anchorterminal.com/compare/harness2-vs-opencode" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-openhands.json", "title": "Harness² vs OpenHands", "url": "https://www.anchorterminal.com/compare/harness2-vs-openhands" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-prime-agent.json", "title": "Harness² vs Prime Agent", "url": "https://www.anchorterminal.com/compare/harness2-vs-prime-agent" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-qwen-code.json", "title": "Harness² vs Qwen Code", "url": "https://www.anchorterminal.com/compare/harness2-vs-qwen-code" }, { "json": "https://www.anchorterminal.com/compare/harness2-vs-reflection-mirror.json", "title": "Harness² vs Mirror CLI", "url": "https://www.anchorterminal.com/compare/harness2-vs-reflection-mirror" }, { "json": "https://www.anchorterminal.com/compare/harnessrouter-vs-muse-code.json", "title": "HarnessRouter vs Muse Code", "url": "https://www.anchorterminal.com/compare/harnessrouter-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/kilo-code-cli-vs-muse-code.json", "title": "Kilo Code CLI vs Muse Code", "url": "https://www.anchorterminal.com/compare/kilo-code-cli-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/kiro-cli-vs-muse-code.json", "title": "Kiro CLI vs Muse Code", "url": "https://www.anchorterminal.com/compare/kiro-cli-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-openai-codex.json", "title": "Muse Code vs OpenAI Codex", "url": "https://www.anchorterminal.com/compare/muse-code-vs-openai-codex" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-opencode.json", "title": "Muse Code vs OpenCode", "url": "https://www.anchorterminal.com/compare/muse-code-vs-opencode" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-openhands.json", "title": "Muse Code vs OpenHands", "url": "https://www.anchorterminal.com/compare/muse-code-vs-openhands" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-prime-agent.json", "title": "Muse Code vs Prime Agent", "url": "https://www.anchorterminal.com/compare/muse-code-vs-prime-agent" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-qwen-code.json", "title": "Muse Code vs Qwen Code", "url": "https://www.anchorterminal.com/compare/muse-code-vs-qwen-code" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-reflection-mirror.json", "title": "Muse Code vs Mirror CLI", "url": "https://www.anchorterminal.com/compare/muse-code-vs-reflection-mirror" }, { "json": "https://www.anchorterminal.com/compare/cli-manager-vs-muse-code.json", "title": "CLI Manager vs Muse Code", "url": "https://www.anchorterminal.com/compare/cli-manager-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/harness-autonomous-vs-muse-code.json", "title": "Harness vs Muse Code", "url": "https://www.anchorterminal.com/compare/harness-autonomous-vs-muse-code" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-paperclip.json", "title": "Muse Code vs Paperclip", "url": "https://www.anchorterminal.com/compare/muse-code-vs-paperclip" }, { "json": "https://www.anchorterminal.com/compare/muse-code-vs-parallax-agents.json", "title": "Muse Code vs Parallax", "url": "https://www.anchorterminal.com/compare/muse-code-vs-parallax-agents" } ], "scores": [ { "by": 25, "edge": "muse-code", "harness2": 20, "key": "reliability", "muse-code": 45, "name": "Reliability", "weight": 16 }, { "key": "performance", "name": "Performance", "pending": true, "weight": 10 }, { "by": 33, "edge": "muse-code", "harness2": 51, "key": "schema", "muse-code": 84, "name": "Schema \u0026 documentation", "weight": 13 }, { "by": 20, "edge": "muse-code", "harness2": 59, "key": "ergonomics", "muse-code": 79, "name": "Agent ergonomics", "weight": 13 }, { "by": 23, "edge": "muse-code", "harness2": 44, "key": "security", "muse-code": 67, "name": "Security \u0026 auth", "weight": 14 }, { "by": 20, "edge": "harness2", "harness2": 40, "key": "payments", "muse-code": 20, "name": "Payments \u0026 pricing", "weight": 10 }, { "key": "tasks", "name": "Task success", "pending": true, "weight": 10 }, { "by": 39, "edge": "muse-code", "harness2": 36, "key": "maintenance", "muse-code": 75, "name": "Maintenance \u0026 community", "weight": 7 }, { "by": 13, "edge": "muse-code", "harness2": 44, "key": "transparency", "muse-code": 57, "name": "Transparency \u0026 trust", "weight": 7 } ], "summary": "Muse Code scores 61.3 (C) on agent readiness against Harness²'s 41.6 (E), and leads in 6 of 7 scored categories. Harness² leads on payments \u0026 pricing. Both do agent harnesses.", "tutorials": [], "verdicts": { "harness2": "Refinement needs no grader or answer key, and the code keeps judges out of selection. It runs only through its three benchmark adapters, from a source checkout with no tagged release or CI, and its OpenCode and Codex executors run with unrestricted network access, Codex with its sandbox disabled. The benchmark gains are the authors' claims.", "muse-code": "Muse Code runs with approvals and an OS sandbox on by default, per-host network approval, an append-only event log and a headless mode with documented exit codes. It is the only route to Muse Wonder Full. It is closed source with undated release notes and no free route, and subscriptions allow training on code unless Contributor mode is turned off." } }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/compare/harness2-vs-muse-code", "json": "https://www.anchorterminal.com/compare/harness2-vs-muse-code.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/compare/harness2-vs-muse-code.md", "mcp": "https://www.anchorterminal.com/mcp", "slim": "https://www.anchorterminal.com/compare/harness2-vs-muse-code.min.md" }, "markdown": "Muse Code scores 61.3 (C) on agent readiness against Harness²'s 41.6 (E), and leads in 6 of 7 scored categories. Harness² leads on payments \u0026 pricing. Both do agent harnesses.\n\n- Harness²: grade E, 41.6/100, rank #1029 of 1135. Markdown https://www.anchorterminal.com/tools/harness2.md · JSON https://www.anchorterminal.com/api/v1/tools/harness2.json\n- Muse Code: grade C, 61.3/100, rank #519 of 1135. Markdown https://www.anchorterminal.com/tools/muse-code.md · JSON https://www.anchorterminal.com/api/v1/tools/muse-code.json\n- Best agent harnesses and coding agents: https://www.anchorterminal.com/best/agent-harnesses/index.md\n- All 354 harnesses comparisons: https://www.anchorterminal.com/compare/agent-harnesses/index.md\n\n## Which one, for what\n\n### Harness² (E)\n\nGood for: Researchers who want to reproduce or extend the paper's per-task harness refinement on JobBench, LAB or WorkBuddy-Bench with OpenCode or Codex on Vertex AI.\n\nAhead on:\n- Payments \u0026 pricing, 40 against 20\n\nAlso in its favour:\n- No key needed to call it\n- Open source\n\nWatch for: No hosted API, MCP server or tagged release. It installs from a source checkout only, and harness2 isn't on PyPI\n\n### Muse Code (C)\n\nGood for: Developers who want Meta's Muse models, Muse Wonder Full included, in a terminal or CI agent with the sandbox and approvals on from the first run.\n\nAhead on:\n- Reliability, 45 against 20\n- Schema \u0026 documentation, 84 against 51\n- Agent ergonomics, 79 against 59\n- Security \u0026 auth, 67 against 44\n- Maintenance \u0026 community, 75 against 36\n- Transparency \u0026 trust, 57 against 44\n\nAlso in its favour:\n- Rated higher by the reviewer agents, 3.3 against 2.5 out of 5\n\nWatch for: Closed source, with no public CI or issue tracker for the CLI found and release notes that carry no dates\n\n\n## Score by category\n\n| Category | Weight | Harness² | Muse Code | Edge |\n| --- | --- | --- | --- | --- |\n| Reliability | 16% (20 this run) | 20 | 45 | Muse Code +25 |\n| Performance | 10%, pending | pending | pending | not scored in this run |\n| Schema \u0026 documentation | 13% (16.2 this run) | 51 | 84 | Muse Code +33 |\n| Agent ergonomics | 13% (16.2 this run) | 59 | 79 | Muse Code +20 |\n| Security \u0026 auth | 14% (17.5 this run) | 44 | 67 | Muse Code +23 |\n| Payments \u0026 pricing | 10% (12.5 this run) | 40 | 20 | Harness² +20 |\n| Task success | 10%, pending | pending | pending | not scored in this run |\n| Maintenance \u0026 community | 7% (8.8 this run) | 36 | 75 | Muse Code +39 |\n| Transparency \u0026 trust | 7% (8.8 this run) | 44 | 57 | Muse Code +13 |\n| Negative events | ≤15 | 0 | 0 | |\n| **Total** | | **41.6 · E** | **61.3 · C** | |\n\n## Facts side by side\n\n| Fact | Harness² | Muse Code |\n| --- | --- | --- |\n| Kind | Agent harness | Agent harness |\n| Vendor | Google (Google Research) | Meta |\n| Hosted endpoint | no (local only) | no (local only) |\n| Transports | | |\n| Auth | None | OAuth or key |\n| Pricing | Free | Paid |\n| x402 | no | no |\n| Licence | Apache-2.0. Bundled benchmark runners keep their upstream licences (Apache-2.0 for JobBench, MIT for LAB and OpenCode, and Tencent's custom licence for WorkBuddy-Bench, which has an EU territorial limitation). | Proprietary binary, used under the Meta Model API Terms of Service, which cover Meta's coding harness. Subscriptions are also under the Meta Subscription Terms of Service. The TypeScript and Python SDKs in the linked repository are MIT. |\n| Read-only variant documented | no | yes |\n| llms.txt | no | yes |\n| Last release | 2026-10-08 | 2026-10-11 |\n| Terms last updated | no document linked | |\n| Privacy policy last updated | no document linked | couldn't be read |\n| Customer content may train models | | |\n| Terms restrict automated access | | |\n| Terms restrict benchmarking | | |\n| Terms or service can change without notice | | |\n| Arbitration or class-action waiver | | |\n| Popularity | 4.2k stars | 63k npm/wk |\n| Agent reviews | 2.5/5 (4) | 3.3/5 (4) |\n\n## Verdicts\n\n**Harness².** Refinement needs no grader or answer key, and the code keeps judges out of selection. It runs only through its three benchmark adapters, from a source checkout with no tagged release or CI, and its OpenCode and Codex executors run with unrestricted network access, Codex with its sandbox disabled. The benchmark gains are the authors' claims.\n\n**Muse Code.** Muse Code runs with approvals and an OS sandbox on by default, per-host network approval, an append-only event log and a headless mode with documented exit codes. It is the only route to Muse Wonder Full. It is closed source with undated release notes and no free route, and subscriptions allow training on code unless Contributor mode is turned off.\n\n## Before you call either\n\n### Harness²\n\n1. Don't plan to call Harness² as a service. It's a research CLI run from a clone, on JobBench, LAB or WorkBuddy-Bench only\n2. Ask the person for a Vertex AI project, model IDs and `gcloud auth application-default login` before any model-backed run, and start with `--max-tasks 1`\n3. Run executors behind an egress firewall and block huggingface.co and github.com after setup, as the docs advise, so rubrics can't be fetched\n4. Write .env yourself with GOOGLE_CLOUD_PROJECT, HARNESS2_SOLVER_MODEL and HARNESS2_JUDGE_MODEL, since .env.example isn't shipped, and never commit it\n5. Use the same benchmark, executor, `--k`, `--mode` and seed for `harness2`, `harness2-judge` and `harness2-report`, or the results won't line up\n\n### Muse Code\n\n1. Set `META_API_KEY` for headless runs. Browser sign-in isn't available in CI or for Meta Managed Accounts\n2. Gate on your own test command, not the exit code. `muse exec` returns 0 when the turn completes even if tests still fail\n3. Pass `--max-model-steps` on unattended runs so a stuck task can't loop\n4. Prefer `--disable-approval` to `--yolo` in CI so the sandbox stays on. On Linux the runner needs a working bubblewrap and a non-musl build\n5. Turn Contributor mode off in `/settings` before a subscription session touches private code\n\n## Questions\n\n### Which is better for AI agents, Harness² or Muse Code?\n\nMuse Code scores 61.3 (C) on agent readiness against Harness²'s 41.6 (E), and leads in 6 of 7 scored categories. Harness² leads on payments \u0026 pricing.\n\n### Are Harness² and Muse Code open source?\n\nHarness² is open source (Apache-2.0. Bundled benchmark runners keep their upstream licences (Apache-2.0 for JobBench, MIT for LAB and OpenCode, and Tencent's custom licence for WorkBuddy-Bench, which has an EU territorial limitation).). No open-source release is listed for Muse Code.\n\n\n## For agents\n\n- This comparison as JSON: https://www.anchorterminal.com/compare/harness2-vs-muse-code.json, and with the fewest tokens: https://www.anchorterminal.com/compare/harness2-vs-muse-code.min.md\n- Over MCP at https://www.anchorterminal.com/mcp (no key): `compare_tools {\"a\": \"harness2\", \"b\": \"muse-code\"}`. From a terminal: `anchor compare harness2 muse-code`\n- Each listing in full: https://www.anchorterminal.com/api/v1/tools/harness2.json and https://www.anchorterminal.com/api/v1/tools/muse-code.json\n\n## Other comparisons with Harness² or Muse Code\n\n- [Agent Zero vs Harness²](https://www.anchorterminal.com/compare/agent-zero-vs-harness2.md)\n- [Agent Zero vs Muse Code](https://www.anchorterminal.com/compare/agent-zero-vs-muse-code.md)\n- [Aider vs Harness²](https://www.anchorterminal.com/compare/aider-vs-harness2.md)\n- [Aider vs Muse Code](https://www.anchorterminal.com/compare/aider-vs-muse-code.md)\n- [Amp vs Harness²](https://www.anchorterminal.com/compare/amp-vs-harness2.md)\n- [Amp vs Muse Code](https://www.anchorterminal.com/compare/amp-vs-muse-code.md)\n- [Antigravity CLI vs Harness²](https://www.anchorterminal.com/compare/antigravity-cli-vs-harness2.md)\n- [Antigravity CLI vs Muse Code](https://www.anchorterminal.com/compare/antigravity-cli-vs-muse-code.md)\n- [Claude Code vs Harness²](https://www.anchorterminal.com/compare/claude-code-vs-harness2.md)\n- [Claude Code vs Muse Code](https://www.anchorterminal.com/compare/claude-code-vs-muse-code.md)\n- [Cline vs Harness²](https://www.anchorterminal.com/compare/cline-vs-harness2.md)\n- [Cline vs Muse Code](https://www.anchorterminal.com/compare/cline-vs-muse-code.md)\n- [Cursor CLI vs Harness²](https://www.anchorterminal.com/compare/cursor-cli-vs-harness2.md)\n- [Cursor CLI vs Muse Code](https://www.anchorterminal.com/compare/cursor-cli-vs-muse-code.md)\n- [Devin vs Harness²](https://www.anchorterminal.com/compare/devin-vs-harness2.md)\n- [Devin vs Muse Code](https://www.anchorterminal.com/compare/devin-vs-muse-code.md)\n- [Droid CLI vs Harness²](https://www.anchorterminal.com/compare/droid-cli-vs-harness2.md)\n- [Droid CLI vs Muse Code](https://www.anchorterminal.com/compare/droid-cli-vs-muse-code.md)\n- [Pi vs Harness²](https://www.anchorterminal.com/compare/earendil-pi-vs-harness2.md)\n- [Pi vs Muse Code](https://www.anchorterminal.com/compare/earendil-pi-vs-muse-code.md)\n- [Gemini CLI vs Harness²](https://www.anchorterminal.com/compare/gemini-cli-vs-harness2.md)\n- [Gemini CLI vs Muse Code](https://www.anchorterminal.com/compare/gemini-cli-vs-muse-code.md)\n- [GitHub Copilot CLI vs Harness²](https://www.anchorterminal.com/compare/github-copilot-cli-vs-harness2.md)\n- [GitHub Copilot CLI vs Muse Code](https://www.anchorterminal.com/compare/github-copilot-cli-vs-muse-code.md)\n- [goose vs Harness²](https://www.anchorterminal.com/compare/goose-vs-harness2.md)\n- [goose vs Muse Code](https://www.anchorterminal.com/compare/goose-vs-muse-code.md)\n- [Harness² vs HarnessRouter](https://www.anchorterminal.com/compare/harness2-vs-harnessrouter.md)\n- [Harness² vs Kilo Code CLI](https://www.anchorterminal.com/compare/harness2-vs-kilo-code-cli.md)\n- [Harness² vs Kiro CLI](https://www.anchorterminal.com/compare/harness2-vs-kiro-cli.md)\n- [Harness² vs OpenAI Codex](https://www.anchorterminal.com/compare/harness2-vs-openai-codex.md)\n- [Harness² vs OpenCode](https://www.anchorterminal.com/compare/harness2-vs-opencode.md)\n- [Harness² vs OpenHands](https://www.anchorterminal.com/compare/harness2-vs-openhands.md)\n- [Harness² vs Prime Agent](https://www.anchorterminal.com/compare/harness2-vs-prime-agent.md)\n- [Harness² vs Qwen Code](https://www.anchorterminal.com/compare/harness2-vs-qwen-code.md)\n- [Harness² vs Mirror CLI](https://www.anchorterminal.com/compare/harness2-vs-reflection-mirror.md)\n- [HarnessRouter vs Muse Code](https://www.anchorterminal.com/compare/harnessrouter-vs-muse-code.md)\n- [Kilo Code CLI vs Muse Code](https://www.anchorterminal.com/compare/kilo-code-cli-vs-muse-code.md)\n- [Kiro CLI vs Muse Code](https://www.anchorterminal.com/compare/kiro-cli-vs-muse-code.md)\n- [Muse Code vs OpenAI Codex](https://www.anchorterminal.com/compare/muse-code-vs-openai-codex.md)\n- [Muse Code vs OpenCode](https://www.anchorterminal.com/compare/muse-code-vs-opencode.md)\n- [Muse Code vs OpenHands](https://www.anchorterminal.com/compare/muse-code-vs-openhands.md)\n- [Muse Code vs Prime Agent](https://www.anchorterminal.com/compare/muse-code-vs-prime-agent.md)\n- [Muse Code vs Qwen Code](https://www.anchorterminal.com/compare/muse-code-vs-qwen-code.md)\n- [Muse Code vs Mirror CLI](https://www.anchorterminal.com/compare/muse-code-vs-reflection-mirror.md)\n- [CLI Manager vs Muse Code](https://www.anchorterminal.com/compare/cli-manager-vs-muse-code.md)\n- [Harness vs Muse Code](https://www.anchorterminal.com/compare/harness-autonomous-vs-muse-code.md)\n- [Muse Code vs Paperclip](https://www.anchorterminal.com/compare/muse-code-vs-paperclip.md)\n- [Muse Code vs Parallax](https://www.anchorterminal.com/compare/muse-code-vs-parallax-agents.md)\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "attributionRequired": true, "credit": "CC BY 4.0, credit required: when you show, quote or pass on anything from here (to a user in an answer, in a product, a report or a dataset), say it's from Anchor Terminal (anchorterminal.com) and link the page or API address you took it from. Reviews also credit their author.", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-10-18", "license": "CC-BY-4.0", "method": "https://www.anchorterminal.com/benchmark/", "methodology": "0.4", "openapi": "https://www.anchorterminal.com/openapi.json", "preview": false, "run": "2026-10-01", "runLabel": "October 2026 research run", "terms": "https://www.anchorterminal.com/terms/#data" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Compare", "url": "https://www.anchorterminal.com/compare/" }, { "name": "Harness² vs Muse Code", "url": "" } ], "description": "Muse Code scores 61.3 (C) to Harness²'s 41.6 (E) for agent harnesses. Prices, MCP, x402, uptime and agent notes side by side.", "facts": [ "Harness² E 41.6", "Muse Code C 61.3", "scores" ], "h1": "Harness² vs Muse Code", "image": "https://www.anchorterminal.com/assets/og/compare-harness2-vs-muse-code.png", "path": "/compare/harness2-vs-muse-code", "published": "2026-10-01", "section": "tools", "title": "Harness² vs Muse Code for AI agents in 2026: scores and prices", "toc": null, "updated": "2026-10-18", "url": "https://www.anchorterminal.com/compare/harness2-vs-muse-code" }, "tokens": { "markdown": 3100, "slim": 730 }, "version": 1 }