# PAEval

> PAEval (Personal Agent Evaluation) compiles independent evaluations, hands-on tests and official documentation for personal AI agents worldwide. Snapshot 9 Oct 2026: 44 products (30 fully profiled), 539 source-backed capability records, 3 independent product-level score sources and 339 cited sources.

Rules that apply to every number on the site:

- Each evaluator's score stays on its own scale; scores from different evaluators are never averaged or converted.
- Documented features (what a vendor says) and measured task performance (what an evaluator tested) are kept separate.
- Unknown is not unsupported: missing evidence is shown as missing, never as zero.
- The PAEval Index is in development. PAEval publishes no score or ranking of its own yet.
- When quoting a score, cite the original evaluator and the date; cite PAEval as "PAEval, snapshot 2026-10-09, https://paeval.com".

## Key pages

- [Overview](https://paeval.com/): All products with every independent score, capability coverage and FAQ
- [Leaderboards](https://paeval.com/leaderboards): Each independent evaluator in full, on its own scale, with samples, intervals and caveats
- [Product directory](https://paeval.com/products): Every tracked product with vendor, region, type and documented pricing
- [Capability matrix](https://paeval.com/capabilities): Documented and observed capabilities by domain
- [Benchmarks & sources](https://paeval.com/benchmarks): Personal-agent evaluations worldwide, classified by what they measure
- [Methodology](https://paeval.com/methodology): Evidence tiers, capability statuses and which scores are eligible
- [About & data downloads](https://paeval.com/about): Scope, independence, downloads, corrections and changelog

## Data

- [products.json](https://paeval.com/data/products.json): products with scores, capability statuses and evidence counts
- [scores.csv](https://paeval.com/data/scores.csv): every third-party score with scale, rank, coverage, interval and URL
- [facts.csv](https://paeval.com/data/facts.csv): all capability records with labels and source URLs
- [sources.csv](https://paeval.com/data/sources.csv): all cited sources with dates
- [llms-full.txt](https://paeval.com/llms-full.txt): every product's summary, scores and capability coverage in one Markdown file

## Products

- [Meta Muse](https://paeval.com/products/muse): Meta Muse is a persistent personal agent from Meta (United States). Independent scores: 9.3/10 on Assistant Benchmark; 20% trusted completion on micro1 PersonalAgentBench (n=10).
- [OpenAI dots](https://paeval.com/products/dots): OpenAI dots is a persistent personal agent from OpenAI (United States). Independent scores: 8.4/10 on Assistant Benchmark.
- [Grok Bot](https://paeval.com/products/grok-bot): Grok Bot is an Individual/Team Agent from xAI (via Cursor) (United States). Independent scores: 7.3/10 on Assistant Benchmark; 30% trusted completion on micro1 PersonalAgentBench (n=10).
- [Claude (including Cowork execution)](https://paeval.com/products/anthropic-claude-agentic): Claude (including Cowork execution) is a general-purpose work agent from Anthropic (United States). Documented capabilities, pricing and cited sources.
- [Manus 2.0 / Manus Studio](https://paeval.com/products/manus-2): Manus 2.0 / Manus Studio is a general-purpose work agent from Manus (Singapore). Documented capabilities, pricing and cited sources.
- [Lindy(Personal Lindy + Teammate)](https://paeval.com/products/lindy-teammate): Lindy(Personal Lindy + Teammate) is an Individual/Team Agent from Lindy (United States). Documented capabilities, pricing and cited sources.
- [Gemini Spark](https://paeval.com/products/gemini-spark): Gemini Spark is a persistent personal agent from Google (United States). Independent scores: 60% trusted completion on micro1 PersonalAgentBench (n=10).
- [Perplexity Computer (Cloud)](https://paeval.com/products/perplexity-computer-cloud): Perplexity Computer (Cloud) is a cloud execution agent from Perplexity (United States). Documented capabilities, pricing and cited sources.
- [Perplexity Personal Computer (local access)](https://paeval.com/products/perplexity-personal-computer): Perplexity Personal Computer (local access) is a cloud + local execution agent from Perplexity (United States). Documented capabilities, pricing and cited sources.
- [Microsoft Copilot Cowork for personal accounts(Preview)](https://paeval.com/products/microsoft-personal-copilot-cowork-preview): Microsoft Copilot Cowork for personal accounts(Preview) is a Personal Execution Agent (Preview) from Microsoft (United States). Documented capabilities, pricing and cited sources.
- [Instinct](https://paeval.com/products/instinct): Instinct is an Universal Personal Execution Agent from Instinct (Spear Street Technology) (United States). Independent scores: 8.5/10 on Assistant Benchmark; 40% trusted completion on micro1 PersonalAgentBench (n=10).
- [Town](https://paeval.com/products/town): Town is a work-based personal agents and team automation from Town.com (United States). Independent scores: 8.3/10 on Assistant Benchmark.
- [Pally](https://paeval.com/products/pally): Pally is a multi-channel universal personal execution agent from Pally Technologies (United States). Independent scores: 8.3/10 on Assistant Benchmark.
- [Ollie](https://paeval.com/products/ollie): Ollie is a home personal assistant from Confabulation Corporation (United States). Independent scores: 8.3/10 on Assistant Benchmark.
- [szn](https://paeval.com/products/szn): szn is an Universal Life Execution Agent from Nomadic Futures (United States). Independent scores: 8.2/10 on Assistant Benchmark.
- [Shuffle](https://paeval.com/products/shuffle): Shuffle is a creation/entertainment companion assistant, including website execution from DNFT (United States). Independent scores: 7.8/10 on Assistant Benchmark.
- [Tomo](https://paeval.com/products/tomo): Tomo is a personal goals/life companionship and execution assistant from Mapo Labs (United States). Independent scores: 7.7/10 on Assistant Benchmark.
- [Asaply](https://paeval.com/products/asaply): Asaply is a local consumption/errand transaction type personal agent from Asaply Inc. (United States). Independent scores: 7.5/10 on Assistant Benchmark.
- [Caddy](https://paeval.com/products/caddy): Caddy is a Message-Based Universal Personal/Home Execution Agent from Caddy Technologies Inc. (United States). Independent scores: 7.3/10 on Assistant Benchmark.
- [Cue by Manus](https://paeval.com/products/manus-cue): Cue by Manus is an independent identity, multi-agent collaborative personal execution agent from Manus (Singapore). Documented capabilities, pricing and cited sources.
- [OpenClaw](https://paeval.com/products/openclaw): OpenClaw is a Self-Hosted Personal Agent from OpenClaw Foundation (Open source / community). Documented capabilities, pricing and cited sources.
- [Hermes Agent](https://paeval.com/products/hermes-agent): Hermes Agent is a Self-Hosted Personal Agent from Nous Research (Open source / community). Documented capabilities, pricing and cited sources.
- [AutoGLM (hosted application)](https://paeval.com/products/autoglm): AutoGLM (hosted application) is a hosted cloud-device agent from Zhipu / Z.ai (China). Documented capabilities, pricing and cited sources.
- [Doubao Phone Assistant (consumer version)](https://paeval.com/products/doubao-phone-assistant): Doubao Phone Assistant (consumer version) is a device-integrated personal agent from ByteDance (China). Documented capabilities, pricing and cited sources.
- [Qwen App (task execution / work assistant)](https://paeval.com/products/qwen-app): Qwen App (task execution / work assistant) is a consumer ecosystem personal agent from Alibaba (China). Documented capabilities, pricing and cited sources.
- [Coze 3.0 / Coze Agent (formerly Coze Space)](https://paeval.com/products/coze-agent): Coze 3.0 / Coze Agent (formerly Coze Space) is a hosted personal work agent from ByteDance (China). Documented capabilities, pricing and cited sources.
- [HONOR YOYO Agent](https://paeval.com/products/honor-yoyo): HONOR YOYO Agent is a device-integrated personal agent from Honor (China). Documented capabilities, pricing and cited sources.
- [MiniMax Agent/MiniMax Code Work Mode](https://paeval.com/products/minimax-work): MiniMax Agent/MiniMax Code Work Mode is a hybrid general-purpose work agent from MiniMax (China). Documented capabilities, pricing and cited sources.
- [QwenWork](https://paeval.com/products/qwenwork): QwenWork is a hybrid personal and team work agent from Alibaba / DingTalk (China). Documented capabilities, pricing and cited sources.
- [WorkBuddy (Tencent)](https://paeval.com/products/workbuddy): WorkBuddy (Tencent) is a desktop personal work agent from Tencent (China). Independent scores: 96.61/100 on SuperCLUE-XClaw (Aug 2026).
- [Wenxin Assistant (task mode)](https://paeval.com/products/baidu-wenxin-assistant): Wenxin Assistant (task mode) is a Claw-style agent (SuperCLUE category) from Baidu (China). Independent scores: 97.62/100 on SuperCLUE-XClaw (Aug 2026).
- [MiMoClaw](https://paeval.com/products/xiaomi-mimoclaw): MiMoClaw is a Claw-style agent (SuperCLUE category) from Xiaomi (China). Independent scores: 96.74/100 on SuperCLUE-XClaw (Aug 2026).
- [KimiClaw](https://paeval.com/products/moonshot-kimiclaw): KimiClaw is a Claw-style agent (SuperCLUE category) from Moonshot AI (China). Independent scores: 96.01/100 on SuperCLUE-XClaw (Aug 2026).
- [MaxClaw](https://paeval.com/products/minimax-maxclaw): MaxClaw is a Claw-style agent (SuperCLUE category) from MiniMax (China). Independent scores: 94.79/100 on SuperCLUE-XClaw (Aug 2026).
- [DuMate](https://paeval.com/products/baidu-dumate): DuMate is a Claw-style agent (SuperCLUE category) from Baidu (China). Independent scores: 93.44/100 on SuperCLUE-XClaw (Aug 2026).
- [AutoClaw](https://paeval.com/products/zhipu-autoclaw): AutoClaw is a Claw-style agent (SuperCLUE category) from Zhipu AI (China). Independent scores: 93.23/100 on SuperCLUE-XClaw (Aug 2026).
- [QwenPaw](https://paeval.com/products/alibaba-qwenpaw): QwenPaw is a Claw-style agent (SuperCLUE category) from Alibaba (China). Independent scores: 92.97/100 on SuperCLUE-XClaw (Aug 2026).
- [ArkClaw](https://paeval.com/products/bytedance-arkclaw): ArkClaw is a Claw-style agent (SuperCLUE category) from ByteDance (China). Independent scores: 92.61/100 on SuperCLUE-XClaw (Aug 2026).
- [StepClaw](https://paeval.com/products/stepfun-stepclaw): StepClaw is a Claw-style agent (SuperCLUE category) from StepFun (China). Independent scores: 91.31/100 on SuperCLUE-XClaw (Aug 2026).
- [ArkClaw-Pro](https://paeval.com/products/bytedance-arkclaw-pro): ArkClaw-Pro is a Claw-style agent (SuperCLUE category) from ByteDance (China). Documented capabilities, pricing and cited sources.
- [QClaw](https://paeval.com/products/tencent-qclaw): QClaw is a Claw-style agent (SuperCLUE category) from Tencent (China). Documented capabilities, pricing and cited sources.
- [ArkClaw-Lite](https://paeval.com/products/bytedance-arkclaw-lite): ArkClaw-Lite is a Claw-style agent (SuperCLUE category) from ByteDance (China). Documented capabilities, pricing and cited sources.
- [CoPaw](https://paeval.com/products/alibaba-copaw): CoPaw is a Claw-style agent (SuperCLUE category) from Alibaba (China). Documented capabilities, pricing and cited sources.
- [DuClaw](https://paeval.com/products/baidu-duclaw): DuClaw is a Claw-style agent (SuperCLUE category) from Baidu (China). Documented capabilities, pricing and cited sources.

## Comparisons, categories and reports

- [What independent tests say about personal AI agents (October 2026)](https://paeval.com/reports/2026-10-independent-evidence): only 23 of 44 personal AI agents have an independent product-level score; the three evaluators barely overlap; 80% of capability records come from vendor documentation.
- [Personal AI agents that work in the background](https://paeval.com/agents/background-agents): evidence on scheduled, triggered and long-running work that continues without you
- [Personal AI agents with cloud or local computers](https://paeval.com/agents/agents-with-computers): evidence on cloud computers, browsers, phones and local machines the agent can operate
- [Personal AI agents that make calls and send messages](https://paeval.com/agents/agents-that-call-and-message): evidence on phone calls, email, messaging and acting under their own identity
- [Personal AI agents with app connectors and tools](https://paeval.com/agents/agents-with-connectors): evidence on connectors, MCP servers, APIs and third-party tools
- [Personal AI agents with long-term memory](https://paeval.com/agents/agents-with-memory): evidence on memory across sessions, personalization and how memory can be edited
- [Personal AI agents in chat apps, voice and other channels](https://paeval.com/agents/agents-in-chat-apps): evidence on the channels and interfaces you use to reach the agent
- [Personal AI agents for research and deliverables](https://paeval.com/agents/agents-for-research-and-documents): evidence on research, documents, slides, data and other finished outputs
- [Personal AI agents that book and buy](https://paeval.com/agents/agents-that-book-and-buy): evidence on purchases, bookings, payments and transaction safeguards
- [Personal AI agents for teams and parallel work](https://paeval.com/agents/agents-for-teams): evidence on sharing, multiple agents and parallel tasks
- [Personal AI agent reliability and recovery](https://paeval.com/agents/agent-reliability): evidence on retries, recovery, monitoring and failure handling
- [Personal AI agent permissions, privacy and security](https://paeval.com/agents/agent-privacy-and-permissions): evidence on permissions, approvals, data handling and security controls
- [Self-hosted and open-source personal AI agents](https://paeval.com/agents/self-hosted)
- [Assistant Benchmark results](https://paeval.com/benchmarks/assistant-benchmark): what the evaluator measures and its published results
- [micro1 PersonalAgentBench results](https://paeval.com/benchmarks/micro1): what the evaluator measures and its published results
- [SuperCLUE-XClaw results](https://paeval.com/benchmarks/superclue-xclaw): what the evaluator measures and its published results
- [Meta Muse vs OpenAI dots](https://paeval.com/compare/muse-vs-dots)
- [Meta Muse vs Grok Bot](https://paeval.com/compare/muse-vs-grok-bot)
- [Meta Muse vs Gemini Spark](https://paeval.com/compare/muse-vs-gemini-spark)
- [Meta Muse vs Instinct](https://paeval.com/compare/muse-vs-instinct)
- [Meta Muse vs Town](https://paeval.com/compare/muse-vs-town)
- [Meta Muse vs Pally](https://paeval.com/compare/muse-vs-pally)
- [Meta Muse vs Ollie](https://paeval.com/compare/muse-vs-ollie)
- [Meta Muse vs szn](https://paeval.com/compare/muse-vs-szn)
- [Meta Muse vs Shuffle](https://paeval.com/compare/muse-vs-shuffle)
- [Meta Muse vs Tomo](https://paeval.com/compare/muse-vs-tomo)
- [Meta Muse vs Asaply](https://paeval.com/compare/muse-vs-asaply)
- [Meta Muse vs Caddy](https://paeval.com/compare/muse-vs-caddy)
- [OpenAI dots vs Grok Bot](https://paeval.com/compare/dots-vs-grok-bot)
- [OpenAI dots vs Gemini Spark](https://paeval.com/compare/dots-vs-gemini-spark)
- [OpenAI dots vs Instinct](https://paeval.com/compare/dots-vs-instinct)
- [OpenAI dots vs Town](https://paeval.com/compare/dots-vs-town)
- [OpenAI dots vs Pally](https://paeval.com/compare/dots-vs-pally)
- [OpenAI dots vs Ollie](https://paeval.com/compare/dots-vs-ollie)
- [OpenAI dots vs szn](https://paeval.com/compare/dots-vs-szn)
- [OpenAI dots vs Shuffle](https://paeval.com/compare/dots-vs-shuffle)
- [OpenAI dots vs Tomo](https://paeval.com/compare/dots-vs-tomo)
- [OpenAI dots vs Asaply](https://paeval.com/compare/dots-vs-asaply)
- [OpenAI dots vs Caddy](https://paeval.com/compare/dots-vs-caddy)
- [Grok Bot vs Lindy(Personal Lindy + Teammate)](https://paeval.com/compare/grok-bot-vs-lindy-teammate)
- [Grok Bot vs Gemini Spark](https://paeval.com/compare/grok-bot-vs-gemini-spark)
- [Grok Bot vs Instinct](https://paeval.com/compare/grok-bot-vs-instinct)
- [Grok Bot vs Town](https://paeval.com/compare/grok-bot-vs-town)
- [Grok Bot vs Pally](https://paeval.com/compare/grok-bot-vs-pally)
- [Grok Bot vs Ollie](https://paeval.com/compare/grok-bot-vs-ollie)
- [Grok Bot vs szn](https://paeval.com/compare/grok-bot-vs-szn)
- [Grok Bot vs Shuffle](https://paeval.com/compare/grok-bot-vs-shuffle)
- [Grok Bot vs Tomo](https://paeval.com/compare/grok-bot-vs-tomo)
- [Grok Bot vs Asaply](https://paeval.com/compare/grok-bot-vs-asaply)
- [Grok Bot vs Caddy](https://paeval.com/compare/grok-bot-vs-caddy)
- [Claude (including Cowork execution) vs Manus 2.0 / Manus Studio](https://paeval.com/compare/anthropic-claude-agentic-vs-manus-2)
- [Gemini Spark vs Instinct](https://paeval.com/compare/gemini-spark-vs-instinct)
- [Perplexity Computer (Cloud) vs Perplexity Personal Computer (local access)](https://paeval.com/compare/perplexity-computer-cloud-vs-perplexity-personal-computer)
- [Instinct vs Town](https://paeval.com/compare/instinct-vs-town)
- [Instinct vs Pally](https://paeval.com/compare/instinct-vs-pally)
- [Instinct vs Ollie](https://paeval.com/compare/instinct-vs-ollie)
- [Instinct vs szn](https://paeval.com/compare/instinct-vs-szn)
- [Instinct vs Shuffle](https://paeval.com/compare/instinct-vs-shuffle)
- [Instinct vs Tomo](https://paeval.com/compare/instinct-vs-tomo)
- [Instinct vs Asaply](https://paeval.com/compare/instinct-vs-asaply)
- [Instinct vs Caddy](https://paeval.com/compare/instinct-vs-caddy)
- [Town vs Pally](https://paeval.com/compare/town-vs-pally)
- [Town vs Ollie](https://paeval.com/compare/town-vs-ollie)
- [Town vs szn](https://paeval.com/compare/town-vs-szn)
- [Town vs Shuffle](https://paeval.com/compare/town-vs-shuffle)
- [Town vs Tomo](https://paeval.com/compare/town-vs-tomo)
- [Town vs Asaply](https://paeval.com/compare/town-vs-asaply)
- [Town vs Caddy](https://paeval.com/compare/town-vs-caddy)
- [Pally vs Ollie](https://paeval.com/compare/pally-vs-ollie)
- [Pally vs szn](https://paeval.com/compare/pally-vs-szn)
- [Pally vs Shuffle](https://paeval.com/compare/pally-vs-shuffle)
- [Pally vs Tomo](https://paeval.com/compare/pally-vs-tomo)
- [Pally vs Asaply](https://paeval.com/compare/pally-vs-asaply)
- [Pally vs Caddy](https://paeval.com/compare/pally-vs-caddy)
- [Ollie vs szn](https://paeval.com/compare/ollie-vs-szn)
- [Ollie vs Shuffle](https://paeval.com/compare/ollie-vs-shuffle)
- [Ollie vs Tomo](https://paeval.com/compare/ollie-vs-tomo)
- [Ollie vs Asaply](https://paeval.com/compare/ollie-vs-asaply)
- [Ollie vs Caddy](https://paeval.com/compare/ollie-vs-caddy)
- [szn vs Shuffle](https://paeval.com/compare/szn-vs-shuffle)
- [szn vs Tomo](https://paeval.com/compare/szn-vs-tomo)
- [szn vs Asaply](https://paeval.com/compare/szn-vs-asaply)
- [szn vs Caddy](https://paeval.com/compare/szn-vs-caddy)
- [Shuffle vs Tomo](https://paeval.com/compare/shuffle-vs-tomo)
- [Shuffle vs Asaply](https://paeval.com/compare/shuffle-vs-asaply)
- [Shuffle vs Caddy](https://paeval.com/compare/shuffle-vs-caddy)
- [Tomo vs Asaply](https://paeval.com/compare/tomo-vs-asaply)
- [Tomo vs Caddy](https://paeval.com/compare/tomo-vs-caddy)
- [Asaply vs Caddy](https://paeval.com/compare/asaply-vs-caddy)
- [OpenClaw vs Hermes Agent](https://paeval.com/compare/openclaw-vs-hermes-agent)
- [Doubao Phone Assistant (consumer version) vs HONOR YOYO Agent](https://paeval.com/compare/doubao-phone-assistant-vs-honor-yoyo)