Zep Cloud

Manages long-term memory for AI agents using a temporal knowledge graph that stores conversations, extracts facts, and indexes them with time-awareness so agents can recall what was true at any point in time.

Evaluated Mar 06, 2026 (0d ago) vcurrent
Homepage ↗ Repo ↗ AI & Machine Learning agent-memory temporal-knowledge-graph conversation-history fact-extraction llm-memory personalization
⚙ Agent Friendliness
62
/ 100
Can an agent use this?
🔒 Security
82
/ 100
Is it safe for agents?
⚡ Reliability
80
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
--
Documentation
85
Error Messages
80
Auth Simplicity
88
Rate Limits
78

🔒 Security

TLS Enforcement
100
Auth Strength
80
Scope Granularity
70
Dep. Hygiene
80
Secret Handling
82

All data encrypted at rest and in transit; project-scoped API keys; no per-session key scoping available.

⚡ Reliability

Uptime/SLA
85
Version Stability
78
Breaking Changes
75
Error Recovery
80
AF Security Reliability

Best When

You need temporally-aware memory where the sequence and timing of facts matters, such as tracking evolving user beliefs or commitments across many sessions.

Avoid When

Your agent is stateless per request or memory requirements are simple key-value lookups without temporal or graph relationships.

Use Cases

  • Build a personal assistant agent that recalls user preferences, past decisions, and evolving facts over months of interaction
  • Power a sales assistant with persistent CRM-style memory of every prospect conversation and extracted commitment
  • Enable a medical information agent to track patient-reported symptoms over time with temporal context
  • Implement dialog state tracking for multi-session customer service bots that remember issue history
  • Feed a research agent structured facts extracted from prior sessions to avoid re-discovering known information

Not For

  • Single-session stateless request/response agents that do not require cross-session persistence
  • Bulk document indexing pipelines — use a purpose-built vector store instead
  • Agents requiring sub-10ms memory retrieval at very high throughput

Interface

REST API
Yes
GraphQL
No
gRPC
No
MCP Server
No
SDK
Yes
Webhooks
No

Authentication

Methods: api_key
OAuth: No Scopes: No

API key passed as Authorization Bearer header; project-level keys scope access to all sessions within a project.

Pricing

Model: freemium
Free tier: Yes
Requires CC: No

Open-source Community Edition (CE) available for self-hosting. Cloud adds managed infrastructure, temporal graph, and enterprise support.

Agent Metadata

Pagination
cursor
Idempotent
Partial
Retry Guidance
Documented

Known Gotchas

  • Fact extraction runs asynchronously after message ingestion — do not assume facts are immediately queryable after adding messages
  • Session IDs must be consistent across the lifetime of a user conversation or history fragments into disconnected sessions
  • The temporal graph is a Cloud-only feature; OSS CE uses simpler vector search without temporal indexing
  • Large conversation histories trigger background summarization that may alter what raw messages are returned
  • Graph traversal for complex relational queries is not exposed in the public API — only semantic search over extracted facts

Alternatives

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Zep Cloud.

$99

Scores are editorial opinions as of 2026-03-06.

5178
Packages Evaluated
26151
Need Evaluation
173
Need Re-evaluation
Community Powered