vinculumsubstrate memory for ai
  • How it worksThe substrate, end to end
  • The fieldParallel sessions, one memory
  • Mission ControlLive fleet telemetry
  • AI failure modesWhat we design against
  • Why it existsThe problem in one page
See the dashboardWatch a live fleet coordinate through one chat window.
Get started
  • QuickstartZero to first grunt
  • ArchitectureHow it's built
  • Self-hostingYour box, your data
  • API referenceEndpoints & tools
Connect your client
  • Claude Code
  • Claude.ai
  • Cursor
  • Cline
  • Zed
The specification
  • OverviewThe open memory spec
  • v0.1 specificationCurrent conformance spec
  • RevisionsHow the spec evolves
  • ContributingPropose a change
The encyclopedia
  • Substrate memoryThe definitional reference
  • vs. vector memory & RAGSimilarity vs. causality
  • The cost of keeping the thinkingCache vs. substrate economics
  • Why LLMs forgetThe intuitive entry point
The library
  • Browse recipesReady-to-run prompts, asks & gates
  • VerifiedCurated & safety-audited
  • CommunityShared by the field
Author
  • Standing OrdersRules every session follows
  • Safe by designFour-tier safety gates
  • Publish a recipeShare a workflow you trust
Hosted plans
  • Free$0 · solo demo, self-host free
  • Substrate$5/mo · hosted memory + dashboard
  • Pro$20/mo · spawn parallel workers
  • Team$100/mo · shared projects, SSO
More
  • EnterpriseOn-prem · SAML · SLA
  • Self-hostFree forever · AGPL v3
  • Compare plansFull feature matrix
Company
  • BlogField notes
  • ChangelogWhat shipped
  • AboutWho's building this
Trust
  • SecurityHow we protect data
  • StatusLive uptime
  • ContactTalk to us
sign in
  1. home›
  2. subprocessors

subprocessors

Every third-party service we use, what data they receive, and where they're based.

last updated July 27, 2026 · source: content/legal/subprocessors.md

Subprocessors

Vinculum uses a small number of third-party services to run the hosted product. This page lists every sub-processor we use, what they handle, and where they are based.


Current subprocessors

ProcessorPurposeLocation
OVHBare-metal server hosting for the application and primary databaseFrance (EU)
AnthropicLLM API — thread title generation, delta classification, entry summaries (Pro/Team tier only)USA
Voyage AIEmbedding API — vector embeddings for semantic search and clustering (Pro/Team tier only)USA
StripePayment processing and subscription managementUSA
CloudflareCDN, edge TLS termination, DNS, and encrypted off-box backup storage (R2)USA / global edge
StalwartTransactional email — account notices, security alerts, billing receiptsUSA

What each processor receives

OVH is the physical home of the Postgres database and application server. OVH staff cannot read application data without compromising the encryption layer — database connections require application credentials; disk encryption is enforced at the host level.

Anthropic receives the text of entries we pass for summarization or classification — only on the Pro/Team tier and only for entries where those features are active. Under Anthropic's commercial API terms, content is not used for model training.

Voyage AI receives entry text for generating vector embeddings — again, only on Pro/Team. Voyage AI's commercial terms prohibit using customer content for training purposes.

Stripe receives your billing identity and manages payment instrument tokenization. We never see your card number; it is entered into a Stripe-hosted iframe and never touches our servers.

Cloudflare terminates TLS at the edge and routes traffic to our origin server. Database backups are stored on Cloudflare R2 in encrypted form (age encryption; Cloudflare sees only ciphertext).

Stalwart handles transactional email only: receipts, security notices, and service alerts. We do not send marketing email through this channel.


Self-hosted instances

If you self-host Vinculum from the open-source repository, none of the above necessarily apply. You choose which integrations to enable via environment variables. Voyage AI and Anthropic features no-op cleanly if their API keys are not configured.


Changes to this list

We will update this page when we add or remove a subprocessor. Material changes — adding any processor that receives content or account data — will be announced via in-product notice and email at least 14 days in advance.


Privacy policies

  • OVH Privacy Policy
  • Anthropic Privacy Policy
  • Voyage AI Privacy Policy
  • Stripe Privacy Policy
  • Cloudflare Privacy Policy

Questions

privacy@vinculum.run

vinculumsubstrate memory for ai
λx.xCopyright 2026 © Whalefall Media, LLC
substrate live

Product

PricingStatusChangelog

Resources

DocsQuickstartMCP IntegrationSelf-hostSubstrate memoryThe StandardGitHub

Company

AboutSecurityPrivacyTermsCookiesAUPSubprocessorsDPAContact

Connect

GitHubX / Twittersteve@vinculum.run