Both projects start from the same idea: your knowledge is Markdown on your disk, and an MCP server lets an agent read and write it. Basic Memory is a released product with a cloud tier and a wide client list. Okto Neuron is a prerelease that derives a graph from the vault and answers with the bytes a claim came from. This page is about the difference, not a ranking.
Both keep memory as Markdown files you can read and serve it to agents over MCP. Basic Memory has a cloud tier, a larger tool surface and more supported clients, and its assistant writes notes for you. Okto Neuron derives a graph from notes you write and returns byte ranges and content hashes with each answer.
Pick Basic Memory if you want wide client coverage, a hosted option, or an assistant that writes the notes.
Pick Okto Neuron if you want each claim traced to bytes in your own files, and an answer that says when the model call failed.
Basic Memory treats the database as a secondary index over standard Markdown and lets the AI write structured notes. Okto Neuron also keeps the Markdown files as the source of record. It builds a graph of claims from them and answers with a pointer to the exact block a claim was read from.
Plain Markdown plus a local SQLite index, no servers required. The AI writes notes in a simple pattern of observations and relations; search returns notes, observations and relations ranked by their best-matching chunks, with the matched chunk text.
Plain Markdown plus a derived graph that can be regenerated from the files. Ingest mints claims from blocks; ask returns text with a synthesis_status, and with include_sources=true, the path, byte range and content hash of every source block.
Competitor cells paraphrase the Basic Memory README, docs and pricing page. Okto Neuron cells were checked against the 0.3.0 source on 2026-09-26 (storage backends, supersession fields, loopback runtime probing, telemetry extra, MCP surface).
| Dimension | Basic Memory | Okto Neuron |
|---|---|---|
| Where data lives | Markdown files on your disk. Optional cloud sync, hosted on Tigris S3, with an air-gapped local mode | Markdown vault on your disk; the graph is derived from it. Nothing is hosted |
| Storage and index | Files plus a local SQLite index (Postgres also supported). Hybrid full-text and vector ranking with FastEmbed | A graph derived from the vault on Okto Grafx by default (Ladybug selectable; Neo4j selectable with the neo4j extra, which the installer does not add). Can be regenerated from the files |
| Agent access | MCP-native, 20+ tools (write_note, read_note, edit_note, search_notes, build_context, and more). Clients listed: Claude Code, Codex, Cursor, VS Code, ChatGPT, Obsidian | MCP with five tools over a bearer-authenticated port; REST and UI on loopback. The installer wires Claude Code |
| Provenance unit | Note, observation and relation permalinks (memory:// URLs). Search returns the matched chunk text, chunks of about 900 characters with about 120 overlap. Note-level checksum | A block: vault-relative path, byte_start, byte_end and a sha256 content_hash, returned with include_sources=true on ask |
| Temporal model | Frontmatter created and modified; recent_activity timeframes. Cloud tier adds snapshots and point-in-time restore | Ingest-time supersession bounded to the document being edited: the old claim gets valid_until and drops from recall, with a supersedes edge. Cross-document corrections are deferred, not applied |
| Extraction | No server-side LLM extraction. The AI client writes structured Markdown; structure emerges from text patterns. Local FastEmbed is for search only | LLM curation at ingest: claims are minted from blocks and a judge confirms corrections. Synthesis on ask reports synthesis_status |
| License | AGPL-3.0 | Elastic License 2.0 from 0.3.0 (source-available). Releases up to 0.2.0 stay Apache 2.0 |
| Hosted option | Basic Memory Cloud at $15/seat/month, Business at $30/seat/month with HIPAA. Local use is free | None |
| Benchmarks | None in the sources reviewed | LoCoMo categories 1-4 from our own harness and judge, on the benchmarks page |
| Maturity | Released product with a paid cloud tier | Prerelease 0.3.0. The upgrade from 0.2.0 has been rehearsed on Linux with a locally built wheel; macOS and Windows are not yet verified on 0.3.0 |
Basic Memory has shipped these, and Okto Neuron has not started them.
Basic Memory Cloud syncs across devices, adds snapshots and point-in-time restore, and offers a Business tier. Okto Neuron has no hosted option and no sync.
Basic Memory lists Claude Code, Codex, Cursor, VS Code, ChatGPT and Obsidian. Okto Neuron's installer wires Claude Code; other clients are unverified.
More than twenty MCP tools covering write, edit, search, context building, schema validation and project management. Okto Neuron ships five tools in a 0.3.0 prerelease.
Basic Memory's whole loop is the client writing structured Markdown back into your vault. Okto Neuron's remember ingests a file or raw text into the graph; it does not author your notes.
Basic Memory's provenance unit is a note, observation or relation permalink plus the matched chunk text and a note-level checksum. Okto Neuron's is a block with byte_start, byte_end and a sha256 content_hash, so a reviewer can open the file and check the exact bytes.
Replace a fact in a note and re-ingest it. The old claim is dated with valid_until, drops from recall, and is linked from the new claim by a supersedes edge. This needs the edit to have replaced the old text and an LLM correction judge to confirm the change, and it only applies inside the document being edited; cross-document corrections are deferred for triage. The correction judge does not sweep the whole vault.
Every ask stamps retrieval.synthesis_status: ok, empty, provider_error, truncated, abnormal_stop or no_llm. Basic Memory does not run an LLM server-side, so this row is a difference in what the tool does, not a gap.
The MCP surface is bearer-authenticated on its own port; REST and UI are credential-free on loopback. Write tools stay loopback-only.
You want the AI to write and maintain the notes, in Markdown, across several clients.
You need sync across devices, hosted backups, or a business tier with HIPAA.
You want a released product with a wide tool surface today.
You want each cited source in an answer to come back with its file path, byte range and content hash.
You want a replaced fact to supersede the old claim on the next ingest, with the old claim kept in the graph and linked to the new one.
You want a graph over your vault, on your machine, with nothing hosted, and you are fine with a prerelease.
No. Our LoCoMo results come from our own harness and a local judge, on categories 1-4 only. Figures other projects publish use their own setup. Ours are on the benchmarks page.
It is source-available. From 0.3.0 the license is the Elastic License 2.0 with an addendum covering SaaS, competing services, internal use and branding. Releases up to 0.2.0 stay under Apache 2.0. The source is on GitHub.
No. It runs on your machine, binds 127.0.0.1 and needs no account.
Every competitor statement on this page comes from one of these pages, read on 2026-09-19, with key rows re-checked on 2026-09-26. If a source has changed since, the page is wrong and we would like to know.
| Claim area | Primary source |
|---|---|
| Positioning, local-first, cloud table, clients | https://raw.githubusercontent.com/basicmachines-co/basic-memory/main/README.md |
| File-first architecture, SQLite or Postgres, FastEmbed | https://docs.basicmemory.com/reference/technical-information |
| MCP tools | https://docs.basicmemory.com/reference/mcp-tools-reference |
| Knowledge format, observations and relations | https://docs.basicmemory.com/concepts/knowledge-format |
| memory:// URLs | https://docs.basicmemory.com/concepts/memory-urls |
| Chunking and search results | https://docs.basicmemory.com/concepts/semantic-search |
| Cloud, snapshots, restore | https://docs.basicmemory.com/cloud/cloud-guide |
| Pricing | https://basicmemory.com/pricing |
Prerelease 0.3.0. Free to run locally. No account required.