| Type | Startup | Startup | Startup |
|---|
| CategoriesDiffers | | | |
|---|
| In one line | The memory layer for AI apps and agents. | Long-term memory for agents, built on temporal knowledge graphs. | Stateful agents with memory, from the team behind MemGPT. |
|---|
| Stage & fundingDiffers | — | — | Seed · $10M seed led by Felicis (2024) |
|---|
| GitHub stars | — | — | — |
|---|
| License | — | — | — |
|---|
| Language | — | — | — |
|---|
| Last push | — | — | — |
|---|
| Works with | — | — | — |
|---|
| Built on | — | — | — |
|---|
| Built on it | — | — | — |
|---|
| Similar toDiffers | | | |
|---|
| People linkedDiffers | 2 | 1 | 2 |
|---|
| Events | 0 | 0 | 0 |
|---|
| Pros | - +MetadataWidely adopted open-source memory layer with about 68k GitHub stars, under Apache-2.0.[1]
- +DocsOffers three deployment paths: a pip/npm library, a self-hosted Docker server with dashboard and auth, or a managed cloud.[1]
- +DocsThe free cloud tier allows 10k adds and 1k retrievals per month, or 50k adds and 10k retrievals with your own LLM key.[1]
| - +MetadataBuilt on Graphiti, its Apache-2.0 temporal knowledge-graph engine with about 32k stars, which supports Neo4j, FalkorDB and Amazon Neptune.[1]
- +DocsCredits are charged only for ingested Episodes; retrieval, storage, users and threads cost 0 credits.[1]
- +DocsEnterprise offers SOC 2 Type II, a HIPAA BAA, bring-your-own-key encryption and bring-your-own-cloud deployment in your VPC.[1]
- +DocsThe free trial tier includes 10,000 credits per month for prototyping.[1]
| - +User reportsIts MemGPT-style memory helps with long-running work: one HN user says it made long-context documentation manageable.[1]
- +MetadataApache-2.0, self-hostable and popular (about 25.3k GitHub stars). The ADE lets you inspect and edit an agent's memory blocks while it runs.[1][2]
- +DocsLetta Code brings its learning memory into a coding-agent harness, with headless and --yolo modes for sandboxes (per Letta staff on HN).[1][2]
- +User reportsOne engineering manager said they would rather build on Letta or Mastra than trust newer personal-agent projects.[1]
|
|---|
| Cons | - −DocsMem0's own README says its LoCoMo scores come from the managed platform, which has proprietary optimizations that the open-source SDK lacks.[1]
- −Independent testIts benchmark claims are disputed: Zep published a rebuttal of how Mem0 ran Zep, practitioners report trouble reproducing the LoCoMo numbers, and a full-context baseline scored about 72.9%.[1][2]
- −DocsGraph memory, SSO, audit logs and on-prem deployment are available only on the custom-priced plan.[1]
- −MetadataLarge maintenance backlog: about 437 open issues and 450 open PRs.[1]
| - −DocsZep Community Edition was deprecated and moved to a legacy folder with no updates, so the full Zep service is now available only as cloud or BYOC.[1][2]
- −DocsPaid self-serve starts at $125/month (Flex), while SOC 2 Type II, the HIPAA BAA and audit logs are listed as Enterprise-only.[1]
- −DocsGraphiti needs an LLM with reliable structured output, and its README warns that other services may cause ingestion failures, especially with smaller models.[1]
- −User reportsThe Mem0/Zep LoCoMo dispute leaves neither vendor's headline benchmark score directly comparable.[1]
| - −Independent testAn independent review says to avoid it for stateless apps or when the main problem is coordinating several agents, because it focuses on memory for a single agent.[1]
- −User reportsMemory extraction that runs on an LLM gets expensive at scale; a competitor estimates $1k-3k per month in API calls at 100K memories a month.[1]
- −User reportsA builder who tried it found that memory architectures like Letta's 'simply move the pain to retrieval'.[1]
|
|---|
| Best for | Teams that want to add user-level memory to chat agents quickly, with an option to self-host. | Agents that need temporally aware, graph-structured memory with enterprise compliance. | Builders of long-lived, stateful single agents who want memory they can inspect and edit, and can self-host. |
|---|
| Watch out for | Benchmark on your own data rather than trusting vendor LoCoMo numbers. | If you need fully self-hosted memory, plan to build on Graphiti yourself. | Budget for the extra LLM calls that memory management makes, and check retrieval quality on your own data. |
|---|
| Links | | | |
|---|