Compare / Meta Muse vs Caddy
Meta Muse vs Caddy
Meta Muse is a persistent personal agent from Meta (United States); Caddy is a Message-Based Universal Personal/Home Execution Agent from Caddy Technologies Inc. (United States). On Assistant Benchmark (1–10, updated 8 Oct 2026), Meta Muse scores 9.3 over 8 dimensions and Caddy scores 7.3 over 9; both were scored on 6 of the same dimensions, shown below. PAEval holds 30 capability records for Meta Muse and 15 for Caddy (as of 9 Oct 2026).
Scores stay on each evaluator's own scale and are never combined. PAEval does not pick a winner.
| Meta Muse | Caddy | |
|---|---|---|
| Vendor | Meta | Caddy Technologies Inc. |
| HQ region | United States | United States |
| Type | Persistent personal agent | Message-Based Universal Personal/Home Execution Agent |
| Pricing (as documented) | Free has a limit; Power is US$20/month, 500M Muse tokens/week; Maximum is US$100/month, 3B/week; automatic monthly renewal. | Public beta, free to start; list site records Free public beta, no standard payment price list. |
| Assistant Benchmark1–10 mean of scored dimensions | 9.3coverage 8/15 | 7.3coverage 9/15 |
| micro1 PersonalAgentBench% of 10 workflows with trusted completion | 20%95% CI 6–51% | Not scored |
| SuperCLUE-XClaw0–100 weighted total of 5 dimensions | Not scored | Not scored |
| Autonomy & background work | Documented | Documented |
| Execution environments | Documented | Partly disclosed / partly tested |
| Calls, messages & identity | Limited or conditional | Documented |
| Connectors & tools | Documented | Documented |
| Memory & personalization | Documented | Partly disclosed / partly tested |
| Channels & interfaces | Documented | Documented |
| Research & deliverables | Documented | Partly disclosed / partly tested |
| Transactions & bookings | Limited or conditional | No record |
| Collaboration & parallelism | Documented | Documented |
| Reliability & recovery | Limited or conditional | Partly disclosed / partly tested |
| Permissions, security & data | Limited or conditional | Documented |
| Capability records | 30 (0 independent) | 15 (1 independent) |
Capability cells show the strongest documented or observed status in each domain; availability is not task success.Add more products →
Assistant Benchmark, dimension by dimension
| Meta Muse | Caddy | |
|---|---|---|
| Carrying out an online task | 9test | 8test |
| Travel booking | 7observed in use | 8test |
| Recommendation quality | 9test | 7test |
| Purchasing a product | 10observed in use | 7test |
| Responding to emails | 10observed in use | 7test |
| Proactive behavior | Not tested | Not tested |
| Running a routine | 10observed in use | Not tested |
| Third-party integrations | 10observed in use | 7test |
| Permissions & privacy | 9observed in use | Not tested |
| Memory | Not tested | 7test |
| Phone calls | Not tested | n/a |
| Multiplayer / groups | Not tested | n/a |
| Chained tasks | Not tested | 7test |
| Proactive restraint | Not tested | Not tested |
| Content creation / games | Not tested | 8test |
Source scores on the 1–10 anchor scale. "Not tested" is missing evidence, not zero.Source ↗