Skip to content
FoxNose vs Directus

A Directus alternative for knowledge & agent retrieval

This page is for teams building structured app content, knowledge bases, and AI/agent retrieval on a managed platform. If you want to self-host on your own SQL database and own the infrastructure, Directus is built for exactly that — this page is for people who want that class of structured backend as a managed service.

The frame: Directus turns your SQL database into an API; FoxNose is the knowledge layer an agent actually retrieves from. And it comes with public flat pricing — no license keys to register, no sales call to get a quote.

Per concern

Assemble it on Directus, or managed on FoxNose

Directus is a database-first API you shape and extend. For a knowledge-and-retrieval workload, several concerns are things you assemble on top of it — the same concerns FoxNose handles as properties of one managed API.

Managed on FoxNoseAssemble it on Directus
Semantic / hybrid searchHybrid semantic search — vector + keyword + filters fused server-side in one call, on every plan including FreeNo native semantic search. Assemble it: pgvector or Pinecone, your own embedding pipeline, a Flow or extension to keep vectors in sync, and your own query path to fuse vectors with filters
Multi-tenancyTenant in the URL, ownership verified per request, and deleting a tenant cascades to their records and embeddingsDIY: a tenant_id field plus role filter rules you have to get right on every collection and endpoint — or one instance per tenant
Referential ownershipStrict reference: parent-child ownership enforced per request at the API, cascade delete including embeddingsSQL foreign-key cascade you configure per relation; no API-level ownership chain and no automatic cascade of derived artifacts like embeddings
Delivery pathDual API: a read-optimized Flux delivery API and a separate authoring Management APIOne dynamic REST + GraphQL surface over your SQL database — the same API reads and writes
Environments & schema promotionEnvironments per project, with versioned schemas and revisionsSeparate installs; promoting schema between them is manual snapshot / diff / apply scripting
License & pricing certaintyPublic flat tiers from $0, usage-metered with a spend cap. No license to register, no sales call for a priceSource-available under the MSCL, with registration keys and two relicenses in three years. Managed Cloud starts at a $99/mo hosting add-on; the paid Team plan is $499/mo; Enterprise is a custom quote

More on the pieces: the agent-native model, the Knowledge MCP, strict reference, and multi-tenant scoping.

Pricing, line by line

What each one costs to start

Compared per line item, not as a single ratio. FoxNose is self-serve from $0; Directus has a free self-hosted Core tier but no free managed-cloud tier — managed Cloud starts at a $99/mo hosting add-on, the paid Team plan is $499/mo, and seats are metered on top.

FoxNose — public self-serve
  • Free — $0 — 10,000 retrievals, 5,000 writes, 0.5 GB data — hybrid search included
  • Starter — $19/mo — 75,000 retrievals, 20,000 writes, 5 GB data
  • Pro — $99/mo — 400,000 retrievals, larger quotas
  • Overage — $0.20 / 1,000 retrievals — Flat on every paid plan; errors are never billed
  • Spend cap, default 2× plan — Self-serve signup — no sales call, no license keys
Directus Cloud
  • Core — $0 — 3 user seats, 25 collections, 5 Flows, advanced RBAC — self-hosted free, or Cloud hosting as a +$99/mo add-on
  • No free managed-cloud tier — Free use is self-hosting; managed Cloud starts at the $99/mo hosting add-on
  • Team — $499/mo (billed annually; $599 monthly) — 10 SSO seats, 50 collections, 20 Flows — extra seats +$50/seat/mo
  • Enterprise — custom contract — SSO (SAML / OIDC), custom scale, offline mode, SLAs
  • Open Innovation Grant — Free self-hosting for orgs under $5M revenue and under 50 employees

See the full breakdown on the FoxNose pricing page — every feature is on every plan; tiers differ by scale, not by capability.

Where Directus fits better

Cases where Directus is the better pick

  • Bring your own SQL database

    This is the Directus headline, and it is a real one: point it at your existing Postgres, MySQL, or SQLite and it wraps that database in a REST and GraphQL API. FoxNose is a managed store — you do not bring a database to it.

  • Self-hosting & infrastructure ownership

    Under the MSCL Innovation Grant you can run Directus on your own servers and own the whole stack. FoxNose is managed-only. If self-hosting or on-prem is a requirement, Directus is built for exactly that.

  • File & asset pipeline with image transforms

    Directus stores files and does on-the-fly image transformations and presets. FoxNose has no asset pipeline — bring a CDN or object store for media.

  • Flows automation & Insights dashboards

    A visual, no-code automation builder and built-in analytics dashboards over your data. FoxNose does not ship a workflow engine or a dashboard product.

  • GraphQL

    Directus exposes a full GraphQL API alongside REST. FoxNose is REST plus an embedded MCP surface — there is no GraphQL endpoint.

  • Generic app backend / BaaS

    For a general-purpose backend behind any app — arbitrary collections, users, files, roles — Directus is a broad BaaS. FoxNose is scoped to structured knowledge and agent retrieval, not a catch-all backend.

The knowledge layer, managed

Typed content, hybrid semantic search, strict-reference multi-tenancy, and an embedded MCP server behind one API — no database to bring, no license to register. Start free with 10,000 retrievals and 5,000 writes, no credit card.

FAQ

Frequently asked questions

Is FoxNose a drop-in replacement for Directus?
No — the center of gravity is different. FoxNose does not let you bring your own SQL database, has no file or asset pipeline, and exposes no GraphQL endpoint. It is a managed knowledge layer built for structured content and agent retrieval: typed collections, hybrid semantic search, strict-reference ownership, and an embedded MCP server behind one API. If your project leans on the parts above, FoxNose is not a substitute for them.
Does Directus have semantic search?
Not natively and not as a first-party feature — built-in semantic search is a long-standing, heavily-upvoted request that remains open and unshipped on their roadmap. Directus gives you rich SQL-style filters and a search parameter that matches root-level string fields with no relevance ranking. Adding semantic search means assembling it yourself: pgvector or Pinecone, your own embedding pipeline, a Flow or extension to keep vectors in sync, and your own query path. On FoxNose, hybrid search — vector, keyword, and filters fused in one call — is on every plan, including Free.
Doesn't Directus have an MCP server too?
Yes. Directus ships a native, permissions-scoped MCP server (v11.13+) with CRUD and schema operations and an audit trail — that is real and worth crediting. The honest contrast is what the tools operate on: Directus MCP is authoring and CRUD over your generic SQL database with keyword-only search, while the FoxNose MCP includes hybrid semantic search tools over a retrieval-optimized store, under the same auth as REST.
What about the Directus license?
Directus relicensed from GPLv3 to BSL 1.1 in 2023, then in April 2026 (with v12) moved to the Monospace Sustainable Core License (MSCL), a source-available license. Free "Innovation Grant" use now requires both under $5M revenue and under 50 employees, software registration keys replace the earlier honor system, there is a four-year conversion back to GPLv3 and a competing-use restriction, and some enterprise features sit behind paid tiers. That is two relicenses in three years. FoxNose is plain SaaS — there is no license to track and no registration key.
Can I self-host FoxNose?
No. FoxNose is managed-only; there is no self-hosted distribution. If self-hosting or owning the infrastructure is a hard requirement, that is exactly the shelf Directus (or Payload) is built for, and you should reach for one of those instead.
How does multi-tenancy compare?
On Directus, multi-tenancy is by convention: a tenant_id field plus role filter rules you have to apply correctly on every collection and endpoint, or one instance per tenant. On FoxNose it is enforced by the data layer — strict-reference collections put the tenant in the URL, ownership is verified per request, and deleting a tenant cascades to their records and embeddings. See the multi-tenant agent memory guide for how that is set up.

Comparison based on publicly available pricing and documentation, checked 2026-07-23.

Directus is a trademark of its respective owner.