Value Proposition & Features
Zep is a Context Engineering Platforms for AI agents that provides persistent, temporally-aware memory as a managed cloud service built on a temporal Context Graph (knowledge graph) architecture.
[1]
[2]
[?] It targets developers and enterprises building AI agents that need sub-200ms retrieval, governed long‑term memory, and compliance-grade provenance rather than ad‑hoc vector stores.
[3]
[2]
[?] [?] Zep’s commercial platform sits on top of its open‑source Graphiti engine, delivering enterprise controls, multi-tenant context graphs, and integrations with agent frameworks and MCP-based tools.
[2]
[4]
[?]
Core product features:
Priority feature list (5–8 bullets):
- Provenance & Lineage Tracking — every derived fact links back to raw episodes, enabling veracity checks, source attribution for LLM‑synthesized answers, and compliance use cases such as GDPR deletion. [10] [?] [?]
- Enterprise Governance & Access Control — attribute‑based access control on graph edges and episodes, user/group policies, and controls for write permissions on MCP connections. [?] [?] [?]
- Strict Ontology & Ingestion Pipeline Support — strict ontology flags and ingestion guides ensure only configured entity/edge types are added, with batch APIs and data‑prep guides for production pipelines. [?] [?]
Product Roadmap / Announcements
As of August 15, 2026,
- 2026-08-13 — Documentation updates: Account Owners can configure enterprise SSO without sharing IdP credentials, custom OpenAI base URLs (e.g., regional hosts), new Strands and Eve integration guides, and documentation for the
zep-strandspackage for Strands. [?] - 2026-08-04 — Memory MCP connections now allow writes by default; administrators can switch to read-only, and connections provisioning users just‑in‑time now recover automatically after user deletion, with improved name display and documentation of how graph search interprets relative calendar phrases. [?]
- 2026-08-03 — A “Memory for Agent Frameworks” landing page was added, the Ecosystem section renamed to Agent Frameworks, and the “Evaluate Zep for Your Use Case” guide rewritten to cover evaluation workflows including ingestion, search scopes, and graph inspection. [?]
- 2026-07-30 — Batch API jobs now send a single completion webhook to avoid duplicate notifications; strict ontology documented for limiting extraction to configured types, and ingestion pipeline guides added. [?] [?]
- 2026-07-29 — Graph node and edge lists can be ordered by valid time (
valid_at), and company email signups receive a credit‑limited 30‑day Enterprise trial before reverting to Free; ABAC constraints are enforced on updates/deletes of graph artifacts. [?] [13] [?] - 2026-07-18 — New “Implement Zep with agents” page under Developer Tools covering the Build with Zep plugin for Claude Code and Codex, plus onboarding and UI tweaks in the graph dialog. [?] [14]
- 2026-07-16 — Blog post detailing benchmarking of NVIDIA Nemotron 3 Embed 1B for agent memory, showing it as Zep’s best‑performing embedding on their production recall workload. [3] [?] [?]
Recent Developments (last 90 days)
- Self-host model change (2026-08-07) — Independent analysis reports that Zep’s Community Edition is deprecated and
getzep/zepnow hosts examples and integrations, making Zep a hosted‑only product; self‑hosting is now limited to building on Graphiti rather than deploying Zep Cloud itself. [?] [?] - Graph traversal & FalkorDB local mode (2026-07-27–08-02) — Graphiti releases add FalkorDB bug fixes and support for fully local operation (embedded FalkorDB, Ollama, no server/cloud LLM), while Zep Cloud exposes context graph traversal APIs for walking neighbors and multi‑hop subgraphs. [?] [?] [1] [?] [2]
- Embedding model upgrade (2026-07-16) — Zep benchmarks multiple embedding models and concludes Nemotron 3 Embed 1B is the strongest tested on its agent‑memory workload, with statistically significant improvements in Recall@10 and a quantized serving path. [3] [?] [?]
- Pricing tier for emerging companies (2026-07-21–08-02) — Zep introduces “Enterprise memory priced for emerging companies” at $13,000 for the first year, targeting companies that have raised $1M–$10M, with more than $40,000 in discounts over a three‑year term. [?] [?] [?] [?]
History and Origin Story
Zep AI was founded in 2023 and participated in Y Combinator’s Winter 2024 batch, building a temporal knowledge‑graph memory layer for AI agents.
[16]
[17]
[?] The founding team centers on Daniel Chalef (founder/CEO), joined by co‑founders including Paul (Pavlo) Paliychuk and Preston Rasmussen, who introduced the project via “Show HN: Zep, Open-Source Graph Memory for AI Apps” alongside the open‑sourcing of Graphiti. [?]
[17]
The company’s origin story is tied to solving “agent context is hard” by publishing the “Zep: A Temporal Knowledge Graph Architecture for Agent Memory” paper and turning its Graphiti engine into widely adopted open‑source infrastructure for agent memory.
[17]
[?]
[18]
Fundraising History
Markdown table (rounds, dates, amounts, lead investor):
| Round | Date | Amount | Lead investor | |
| Pre-Seed | 2024-03-05 | $500k | Y Combinator | [?] |
| Seed | No reliable source found | – | – |
Total funding: Approximately $500k reported pre‑seed funding. [?]
Investors (alphabetical, from known data):
- Y Combinator [?]
Notable Team Members
Daniel Chalef (Founder & CEO) — Daniel Chalef is the founder and CEO of Zep AI, described as a two‑time founder and former head of ML at SparkPost, and he represents the company publicly in podcasts and conference talks on temporal knowledge graphs and agent memory.
[16]
[17]
[?] [?] His role includes leading Zep’s build‑in‑public motion, presenting technical work such as provenance for LLM‑built knowledge graphs and the Zep memory architecture. [?] [?] [?] [?]
Paul (Pavlo) Paliychuk (Co‑founder) — Systemaic’s teardown cites Paul (Pavlo Paliychuk) as a co‑founder alongside Daniel, part of the founding team that launched Zep’s open‑source and commercial agent‑memory platform. [?]
Preston Rasmussen (Co‑founder) — The same teardown notes Preston Rasmussen in the early “Show HN” introduction (“Daniel, Paul, Travis, and Preston from Zep”), indicating his role as a co‑founder involved in engineering and distribution around Graphiti and Zep’s memory layer. [?]
Market Sizing
Category, Market Size, and Category Growth
Zep operates in the AI agent memory systems infrastructure category and the broader agentic AI orchestration and memory systems and RAG/context tooling ecosystems.
[5]
[19]
[?] [?] One market research report estimates the global AI Agent Memory Systems Infrastructure market at $1.2 billion in 2025 , projected to reach $18.9 billion by 2034 at a 62.0% CAGR . [?] Another report places Agentic AI Orchestration and Memory Systems at $10.9 billion in 2026 , rising to $210.5 billion by 2035 with a 38.9% CAGR , illustrating rapid growth in the stack Zep participates in. [?] Retrieval‑augmented generation markets are also forecast to grow from around $1.94 billion in 2025 to $9.86 billion by 2030 (38.4% CAGR), indicating a strong adjacent demand for context and memory tooling. [?] [?] [?]
Pricing
Zep’s published and reported pricing:
| Tier / Program | Price / Structure | Notes |
| Free | $0; 10,000 credits/month, no rollover | Entry tier for small teams. [?] |
| Flex | $1,250/year (≈$104/month billed annually) with 50,000 monthly credits | Self‑service for individual developers. [?] [?] [?] |
| Flex Plus | $3,750/year (≈$312/month billed annually) with 200,000 monthly credits | Higher‑volume self‑service. [?] [?] [?] [?] |
| Emerging Companies Program | $13,000 for first year, >$40,000 in discounts over three‑year term; 12‑month commitment prepaid annually | For companies with $1M–$10M funding, Enterprise controls at lower price. [?] [?] [?] [?] |
| Enterprise | Custom; usage‑based credits, negotiated rates, deployment options and SLAs | SOC 2 Type II, HIPAA BAA, BYOK and enterprise features. [?] [?] [?] |
Other third‑party sources mention starting pricing “from $25/mo” or “from ~$125/month,” but these are secondary summaries; the most detailed tier descriptions are from Zep’s docs, Systemaic, and HydraDB. [?] [?] [?] [?] [?]
Revenue Trajectory Estimates
Systemaic’s teardown states that Zep’s revenue is not publicly disclosed and offers a qualitative estimate that, given enterprise contracts alongside $104–$312/month self‑serve tiers and a lean team, revenue is “plausibly in the low‑to‑mid six‑figure annual range,” emphasizing this is an estimate, not a confirmed figure. [?] Y Combinator’s job listing for Zep notes that the company is seed‑stage with 50% month‑over‑month ARR growth and 240+ customers including Fortune 500s , but does not disclose exact ARR.
[17]
Competitive Landscape
Who it’s for, who it’s not for
Zep is for developers and enterprises building AI agents that require persistent, temporally-aware memory with strong governance, compliance, and provenance, especially where understanding how facts change over time and across conversations is critical.
[3]
[2]
[5]
[19]
[17]
It is particularly suited to production AI agents, copilots, and enterprise applications needing sub‑200ms retrieval, SOC 2/HIPAA controls, attribute‑based access control, and multi‑tenant context graphs served inside VPCs or controlled cloud deployments.
[2]
[17]
[?] [?]
[19]
Zep is less suited for teams that require fully self‑hosted managed memory products with no external SaaS component, since independent analysis indicates the Community Edition is deprecated and Zep is now a hosted‑only product, leaving self‑hosting to Graphiti plus custom plumbing. [?] [?]
[20]
[?] It is also not ideal for simple, lightweight memory where a flat key‑value store or basic vector DB suffices, or for teams whose budget or complexity requirements favor lower‑cost, simpler memory APIs without temporal graphs and governance overhead.
[21]
[22]
[23]
[?]
Viable Alternatives
Competitor Table
| Competitor | Description |
| Mem0 | Agent memory platform offering a free managed tier and Apache‑licensed self‑hostable package, focusing on vector-based memory and simple integration, widely adopted and integrated into major clouds. [22] [21] [?] [?] |
| Letta (MemGPT) | Open‑source agent framework with OS‑style tiered memory, enabling hierarchical, agent‑managed context without a separate memory SaaS; strong fit for self‑hosters. [24] [5] [22] |
| Supermemory | Hosted and self‑hostable vector‑graph memory API emphasizing ease of use and fast universal ingest, positioned as a straightforward hosted memory layer. [5] [22] [25] |
| Cognee | Agent memory/knowledge‑graph system combining graphs and embeddings, suitable for multi‑agent and data‑heavy knowledge pipelines with configurable storage backends. [26] [5] [27] |
| HydraDB | Graph‑native context layer and AI graph database marketed as an alternative to Mem0 and Zep, designed for production AI agents and enterprise applications requiring complex state tracking. [23] [28] [?] |