ZeroDB vs Hyperspell
Hyperspell is a connector-driven memory and context layer for AI agents — it indexes enterprise tools like Slack, Gmail, and Notion into a permission-aware knowledge graph. Here's what's actually different, based on Hyperspell's own public docs and site, fetched on 2026-11-04.
| Dimension | ZeroDB | Hyperspell | Checked |
|---|---|---|---|
| Pre-built connectors | Five real, wired connectors — Slack, Gmail, Google Drive, Notion, and GitHub — synced via a Celery background task against a provider registry. | 50+ pre-built connectors (Slack, Gmail, Notion, Linear, Jira, Salesforce, HubSpot, GitHub, and more), positioned as the core product surface. | hyperspell.com homepage, fetched 2026-11-04 — lists 50+ connectors as a headline feature. |
| Memory model | Four explicit tiers — working, episodic, semantic, procedural — each memory tagged with memory_type, with promote() to move a memory up a tier, plus a relate()/graph() knowledge graph over entities. | A continuously-synthesized "company knowledge model" built from connected sources, queried via a natural-language POST /memories/query endpoint — not split into named cognitive tiers. | docs.hyperspell.com/core/integration, fetched 2026-11-04 — no working/episodic/semantic/procedural terminology found; describes a single synthesized memory layer. |
| Permission model | Namespace-scoped by default (global / project:<id> / session:<id>) — every write and read is explicitly scoped, with no automatic inheritance from a connected third-party tool's ACLs. | Permission-aware: automatically inherits each connected tool's own OAuth/ACL permissions (e.g. a Slack channel's membership) so agent queries only surface what that user could already see. | hyperspell.com homepage, fetched 2026-11-04 — advertises automatic OAuth permission inheritance across connected tools. |
| Getting started | Single unauthenticated POST provisions a project + API key — no signup form (npx zerodb-cli init), confirmed by a real run this session. | Dashboard-issued API key, used as an Authorization: Bearer header; a free developer tier exists but requires account creation. | docs.hyperspell.com/core/integration, fetched 2026-11-04 — describes dashboard-issued bearer tokens, not zero-auth provisioning. |
| Compliance | See ZeroDB's own security documentation for current certifications and data-handling policy. | SOC 2 certified, GDPR compliant; states customer data is not used to train foundation models. | hyperspell.com homepage, fetched 2026-11-04. |
| Pricing | Hyperspell's homepage surfaced no dollar amounts (CTA is “Book a demo”), while secondary sources describe a free developer tier, usage-based Pro tier, and custom Enterprise pricing without exact figures. We did not independently verify specific prices on docs.hyperspell.com/pricing in this session, so this is left as an open question rather than a guessed number. | ||
| Retrieval quality across 50+ connected sources | Hyperspell's differentiator is breadth of connected enterprise sources feeding one synthesized knowledge graph. We did not provision a Hyperspell account and run a side-by-side multi-source retrieval benchmark against ZeroDB in this session — left as an open question rather than an unverified claim. | ||
What this page does and doesn't claim
Hyperspell and ZeroDB solve adjacent but not identical problems: Hyperspell is built around ingesting a company's existing SaaS tools at scale (50+ connectors) into one permission-aware knowledge graph for agents; ZeroDB's memory layer is built around explicit cognitive tiers and a knowledge graph an agent writes to directly, with a much smaller set of built-in source connectors today.
Every Hyperspell-side cell above cites the specific page fetched and the date. Nothing on this page asserts a superlative ranking ("#1", "best") — it states specific, checkable differences only, including where ZeroDB currently has fewer built-in connectors than Hyperspell.