Compare / vs Mem0
Evidence-bound memoryvs drop-in API.
Mem0 is a fast, popular drop-in memory API, and it is easier to get running than Zephr is. What it does not do is bind a memory to the code it came from, or refuse to answer when that code has moved. The two products are not interchangeable, and picking between them is mostly a question of which risk you are managing.
Same market, different shape.
Mem0 optimises for fast, global user memory. Zephr optimises for scoped, provenance-backed memory that survives a client switch. Neither column is a scorecard — they are different answers to different questions.
| Feature | Mem0 | Zephr |
|---|---|---|
| Memory scope | User memory (global) | Per-worktree, per-session, per-project |
| Authorization | None (open retrieval) | Trust Firewall — authorization before retrieval |
| Provenance | Basic (source tag) | Full chain: source → episode → worktree → session → authorization |
| Cross-client | API (generic) | MCP-native — works across all connected agents |
| Deployment | OSS self-host + managed platform | Local-first SQLite; managed cloud planned |
| Graph memory | Optional / often platform-tier | Evidence graph + scope — not a generic entity graph product |
| Open-core | Yes (Apache-2.0 API) | Yes — local SQLite, no account required |
Public positioning snapshot, re-checked August 2026 — not live data. Mem0 is actively developed and may add comparable capabilities; this is a dated view, not a forecast.
Checkable capabilities, including where Zephr is partial.
A matrix that came out all-green for Zephr would mean the rows were chosen to flatter us. Drop-in API speed is a real Mem0 win; coding scope, authorization-before-retrieval, and signed handoff are the Zephr bets.
| Scope | Drop-in API | Coding scope | Auth before retrieve | Signed handoff |
|---|---|---|---|---|
| Mem0 | ●supported | ◐partial | ○absent | ○absent |
| Zephrthis page | ◐partial | ●supported | ●supported | ●supported |
Scope lens · public positioning re-checked 2026-08-12 · ADR-Z-16 · packages/core/src/admission.ts · packages/mcp-server/src/transport.ts
verdicts are testable and dated — no scores, no winners
Mem0 is faster to adopt and simpler to call.
Zephr is not better at everything, and pretending otherwise would undermine the one thing this product is for. Mem0's strengths below are real, and for a large number of teams they are the deciding factor.
Fast onboarding
Install, authenticate, add, search. Four steps to a working memory, against cloning a repo and building from source.
Larger community
Broader MCP and plugin distribution, more integrations in the wild, and more people who have already hit your problem.
Simpler surface
A small API for generic user memory. If you do not need scope binding, that simplicity is a genuine feature rather than a gap.
Proven traction
Large open-source adoption and a managed platform path — a real signal about maintenance and distribution. Star counts move; treat any number as a dated snapshot, not a permanent ranking.
Public positioning re-checked August 2026. Star counts move; we do not treat a single GitHub number as a permanent ranking.
Where Zephr adds value.
Zephr is not a faster Mem0. It is a different layer — trust, scope, and provenance for coding workflows — and it is worth more the more agents and branches you move between.
Evidence-bound memory
A claim is stored with the files, commits, and line ranges it rests on — not as an opaque vector whose origin cannot be reconstructed.
Worktree and session provenance
Memory binds to the coding context it came from, so a decision made on one branch does not silently answer a question about another.
Trust Firewall
Authorization is evaluated before retrieval rather than filtering afterwards, which is the difference between a policy and a preference.
Coding-specific continuity
Branch overlays and signed handoff packets, built for moving between agents mid-task rather than for remembering user preferences.
Choose Mem0 or Zephr?
The choice comes down to which failure you are more worried about: retrieval being slow to set up, or retrieval being confidently wrong.
Neither failure mode is exotic: onboarding friction is a first-week cost, a confidently wrong answer is a live risk every week after. Which one you are absorbing decides the category.
Choose Mem0 if…
You need global user memory, fast onboarding, and a broad plugin ecosystem — and you can accept basic source tagging without a provenance chain or retrieval authorization.
Choose Zephr if…
You work across multiple coding agents, need per-worktree memory scope, and want every retrieval to prove authorization and carry a receipt you can inspect afterwards.
See how Zephr enforces trust.
The trust model covers authorization before retrieval, scope binding, and evidence receipts — the three things this comparison keeps returning to.