Type: comparison · sources read 2026-09-19, key rows re-checked 2026-09-26

Okto Neuron vs mem0, Graphiti, Basic Memory and the MCP memory server. Where each one is ahead.

"Agent memory" now covers projects that do quite different jobs. Okto Neuron is a prerelease with five MCP tools, and this page says where the others are ahead. Each comparison below is built from the other project's own README, docs and pricing page as read on 2026-09-19 (key rows re-checked on 2026-09-26), names the places where that project is ahead of Okto Neuron, and ends with who should pick which. Our own LoCoMo results live on the benchmarks page, not here: they come from our harness and judge, so they do not compare with the figures these projects publish about themselves.

Comparison · 01 Okto Neuron vs Basic Memory

Both keep knowledge as Markdown on your disk and expose it over MCP. Basic Memory is a released product with 20+ tools, a wide client list and a cloud tier; its provenance unit is a note or chunk permalink. Okto Neuron's is a byte range and a content hash inside a derived graph.

Markdown-first Chunk vs byte range Cloud tier AGPL vs ELv2
Ahead: hosted tier, clients, tool surface, maturity

Pick Okto Neuron if you want each answer's claims to point at byte ranges in files you wrote.

Read the comparison →
Comparison · 02 Okto Neuron vs Graphiti (Zep)

Graphiti builds temporal knowledge graphs on Neo4j, FalkorDB, Kuzu or Amazon Neptune, with bi-temporal edges and point-in-time queries that Okto Neuron does not have. Okto Neuron keeps your Markdown files as the source of record and traces each claim to bytes rather than to an episode.

Bi-temporal edges Episode vs block Several graph backends Apache 2.0 vs ELv2
Ahead: temporal model, managed service through Zep, backends

Pick Okto Neuron if your source of truth is a folder of Markdown notes you write, and you want claims traced to those bytes.

Read the comparison →
Comparison · 03 Okto Neuron vs mem0

mem0 is a memory layer that learns a user's preferences from conversations, shipped as a library, a self-hosted server and a hosted platform with SDKs and benchmarks. Okto Neuron answers from notes you wrote and returns the bytes it read. The overlap is smaller than the category name suggests.

Memory vs claims Hosted MCP ADD-only OSS Apache 2.0 vs ELv2
Ahead: platform, SDKs, client reach, benchmark coverage, maturity

Pick Okto Neuron if the memory is your own notes, answered locally with a pointer to the source.

Read the comparison →
Comparison · 04 Okto Neuron vs the MCP Knowledge Graph Memory server

The reference memory server from the Model Context Protocol servers repository stores entities, relations and observations your agent writes, in a JSONL file, with nine tools and no LLM. Okto Neuron extracts claims from notes you write and keeps a byte-range pointer on each one.

Agent-written vs extracted JSONL vs graph store MIT vs ELv2
Ahead: simpler setup, no model to run, reference implementation

Pick Okto Neuron if you want the graph derived from notes you already write, with citations back to the source.

Read the comparison →

Spotted a claim about one of these projects that is out of date or wrong? Each page lists the exact source it was read from. Tell us and we will fix the page.

Send a correction ↗

Point it at your notes. Check the bytes yourself.

Prerelease 0.3.0. Free to run locally. No account required.