Zephr

Compare / vs Zep

Open-core coding continuityvs enterprise graph memory.

Zep and Graphiti build a powerful temporal graph for enterprise governance. Zephr builds portable, scoped continuity for coding agents. The overlap is provenance, and it is a genuine overlap — the divergence is deployment model, primary abstraction, and who the product is for.

Snapshot · August 2026ADR-Z-16re-checked 2026-08-12
Shared ground

Both start from provenance.

Zephr and Zep/Graphiti agree on the premise that most of this market does not: memory without provenance is unsafe. Both trace beliefs back to source episodes, code references, and session context.

  • Provenance-first memory

    Both products treat an unsourced memory as a liability. Every derived belief traces back to the episodes that produced it.

  • Source-episode lineage

    Code references, session context, and worktree metadata are carried with the record rather than reconstructed on demand.

  • Temporal truth models

    Both encode that a fact was true over an interval, not merely that it is true now — which is what makes historical questions answerable.

  • Governance primitives

    Roles, policies, and audit are treated as core concerns rather than as an enterprise add-on bolted onto a retrieval layer.

Four axes

Shared ground, different primary product.

Graphiti is a real temporal graph engine. Zep Community Edition is no longer maintained — the live split is Graphiti OSS versus Zep Cloud. Zephr is partial on generic temporal graphs and managed cloud; strong on coding handoff and local-first private mode.

BoundaryTemporal graphCoding handoffLocal-firstManaged cloud
Zep / Graphitisupportedpartialpartialsupported
Zephrthis pagepartialsupportedsupportedpartial

Boundary lens · public positioning re-checked 2026-08-12 · Zep OSS strategy note · ADR-Z-16

supported partial absent

verdicts are testable and dated — no scores, no winners

Zep announced a new open-source direction concentrating on Graphiti and stopping Community Edition maintenance. Compare the products you can actually run today — not a retired CE install path.

Divergence

Enterprise graph vs coding continuity.

The two products diverge on deployment model, primary abstraction, and the workflows they were built to optimise. None of these rows is a defect on either side — they are different design centres.

DimensionZep / GraphitiZephr
Primary modelTemporal knowledge graphEvidence graph with worktree overlays
DeploymentEnterprise-focused offeringLocal-first SQLite, managed cloud planned
Open-coreNo — enterprise licensingYes — local SQLite, no account required
Coding continuityGeneral-purpose graph memoryWorktree and branch deltas, signed handoff packets
Cross-clientClosed enterprise integrationMCP-native across connected agents

Public positioning snapshot, re-checked August 2026 — not live data. Rows will move as both products develop; Zep CE is historical, Graphiti is the live OSS engine.

Where Zep wins

Zep's graph model is more mature for enterprise governance.

Zephr does not claim to out-graph Zep. If your workload is enterprise RBAC, temporal facts, and governance audit across a large organization, Zep/Graphiti is the more mature choice today and we would rather say so here than have you find out later.

Zep / GraphitiStronger here

Where the graph model leads

  • Graphiti: open-source temporal knowledge graph with validity windows on facts
  • Zep Cloud: managed path for production graph memory and governance
  • Source-episode lineage designed for organizational audit workloads
ZephrStronger here

Where the continuity layer leads

  • Coding-specific continuity: worktrees, branch overlays, signed handoff
  • Local-first SQLite path with no account required for private mode
  • Evidence receipts as durable, inspectable objects under Guard

Comparison reflects public positioning as of August 2026 (Graphiti OSS + Zep Cloud; CE sunset).

Decision framework

Choose Zep or Zephr?

Pick the product whose primary abstraction matches your primary workflow. A graph you do not query is overhead, and a continuity layer you never switch tools inside is too.

The overlap is provenance, so the decision is about the shape around it: who runs the deployment, which abstraction your queries speak, and how many tools you switch between.

Choose Zep / Graphiti if…

You need temporal graph queries, enterprise RBAC, and governance audit across a large organization — and an enterprise-focused deployment model is a fit rather than an obstacle.

Choose Zephr if…

You want portable coding continuity, local-first deployment, and evidence receipts that survive moving between Claude Code, Cursor, Copilot, and other agents mid-task.

See how continuity works.

The golden path covers install, scope admission, provenance-backed recall, and signed handoff — the parts of the story a graph model alone does not answer.