Friend vs Granola
Friend vs Granola: side-by-side comparison of two voice-first / wearable ai memory systems — architecture, taxonomy, license, pricing, MCP/A2A support, and direct edges.
Cost & capability
| Friend | Granola | |
|---|---|---|
| Capability band | entry | entry |
| Capability composite | 28 | 40 |
| Cost tier | mid | free |
| $/Mtok input | — | 0 |
| $/Mtok output | — | 0 |
| Use cases | Latency Sensitive, Memory Augmented Chat, Long Running Session | Analytical Summarization, Memory Augmented Chat |
Where they differ (9)
Rows where both sides have data and the values disagree — the shortlist of dimensions that actually distinguish these two systems.
| Friend | Granola | |
|---|---|---|
| Capability composite | 28 | 40 |
| Cost tier | mid | free |
| Use cases | Latency Sensitive, Memory Augmented Chat, Long Running Session | Analytical Summarization, Memory Augmented Chat |
| Type | Always-on AI companion pendant | Mac/iOS meeting notepad — system-audio tap |
| Created | 2024 (founded 2024; $2.5M seed Jul 2024; began shipping summer 2025) | 2023-05 |
| Pricing | Hardware-bundled | Free + paid |
| Funding | $10M total (self-raised $2.5M seed + additional $7.5M; backers: Caffeinated Capital, Z Fellows, Perplexity CEO Arav… | $192M total $1.5B val Series C · 2026-03 |
| Multi-tenancy | hard-isolation (per device) | hard-isolation |
| OpenTelemetry | searched not found | no — consumer product |
At a glance
| Friend | Granola | |
|---|---|---|
| Section | Voice-first / wearable AI memory | Voice-first / wearable AI memory |
| Tier | T1 | T1 |
| Type | Always-on AI companion pendant | Mac/iOS meeting notepad — system-audio tap |
| Created | 2024 (founded 2024; $2.5M seed Jul 2024; began shipping summer 2025) | 2023-05 |
| Pricing | Hardware-bundled | Free + paid |
| Funding | $10M total (self-raised $2.5M seed + additional $7.5M; backers: Caffeinated Capital, Z Fellows, Perplexity CEO Arav… | $192M total $1.5B val Series C · 2026-03 |
| Backend storage | custom | custom |
| Deployment | Managed-only | Managed-only |
| API surface | — | searched not found |
| Embedding | locked | locked |
| Multi-tenancy | hard-isolation (per device) | hard-isolation |
| MCP | searched not found | searched not found |
| A2A | searched not found | searched not found |
| OpenTelemetry | searched not found | no — consumer product |
| Optimised for | always-on capture + lifelog timeline + voice UX | always-on capture + lifelog timeline + voice UX |
| Anti-fit | not for enterprise compliance-regulated environments without explicit consent flows | not for enterprise compliance-regulated environments without explicit consent flows |
Taxonomy
| Axis | Friend | Granola |
|---|---|---|
| storage | vector | file |
| retrieval | similarity | extraction-pull |
| persistence | lifelong | long-term |
| update | append-only | append-only |
| unit | episode | episode |
| governance | opaque | user-controllable |
| conflict | append | append |
Pros & cons
Friend
Pros: Companionship framing rather than productivity — addresses a different need than the rest of the wearable category.
Cons: Consumer reaction has been mixed; cloud LLM backend with limited privacy disclosure.
Granola
Pros: Captures system audio without joining as a meeting bot — works for any meeting tool without admin rules; enterprise tier opts out of model training.
Cons: Mac/iOS only — no Windows/Linux story; cloud processing despite local-audio claim.