Pay for what your agents use
Four usage axes — retrievals, writes, data storage and vector storage. Every plan includes the full platform. Start free, upgrade when your allowance runs out.
Free
Prototypes and evaluation.
$0/ mo
Free forever
No credit card required
- Retrievals
- 10,000 / mo
- Writes
- 5,000 / mo
- Data storage
- 0.5 GB
- Vector storage
- 0.1 GB
Hard stop at the allowance. No card on file — a surprise bill is impossible.
Starter
A single agent in production.
$19/ mo
Billed monthly
Upgrade anytime from Free
- Retrievals
- 75,000 / mo
- Writes
- 20,000 / mo
- Data storage
- 5 GB
- Vector storage
- 1 GB
Run past an allowance and nothing breaks: pay as you go at flat per-unit rates, with a spend cap you control.
Pro
Production workloads at scale.
$99/ mo
Billed monthly
Upgrade anytime from Free
- Retrievals
- 400,000 / mo
- Writes
- 100,000 / mo
- Data storage
- 20 GB
- Vector storage
- 4 GB
Run past an allowance and nothing breaks: pay as you go at flat per-unit rates, with a spend cap you control.
Enterprise
Custom volume and terms.
Let's talk
Volume-based pricing
Defaults start at Pro limits
- Retrievals
- Custom
- Writes
- Custom
- Data storage
- Custom
- Vector storage
- Custom
Allowances and overage terms set per contract.
Every plan includes hybrid search, auto-generated Flux APIs, an embedded MCP server on each API, schema and revision versioning, isolated environments, and EU-hosted embeddings.
Metering you can predict
Reads and writes are metered separately. The rules are the same on every plan, so you can predict a bill before you send a request.
Retrievals count 1:1
A retrieval is a single read from a Flux API — a GET, a search, a schema or router call — or one MCP tools/call. A search that returns 100 results is still one retrieval. MCP handshake calls (initialize, tools/list, ping) are free.
Writes are weighted by size
Publishing content generates embeddings, and larger content costs more to embed. One write covers content up to roughly 500 embedded tokens. Beyond that, writes scale in proportion to the amount of text embedded.
Why a single write can cost more than one
We weight writes so a short record and a long document don't cost the same to embed. The formula is straightforward:
- —A record with ~400 embedded tokens counts as 1 write.
- —A document with ~1,500 embedded tokens counts as 3 writes.
Embedding is the expensive part of a write, so the cost tracks the embedding work done — a heavy document isn't priced like a tweet. Only the text you mark as searchable is embedded, so trimming what you index directly lowers write cost.
Only successful requests are billed
Retrievals and writes are counted for 2xx and 3xx responses only. Client and server errors — 4xx, 5xx, rate-limited 429, and OPTIONS preflight — never touch your allowance. A failed request costs you nothing.
Your agents keep working. The bill won't surprise you.
On a paid plan, going over an allowance doesn't stop your traffic — you pay the same flat rate for the extra usage, with no penalty multiplier. A spend cap sets the ceiling.
Overage rates
Same rate on every paid plan. Charged after the cycle closes.
- Retrievals
- $0.20per 1,000
- Writes
- $0.50per 1,000
- Data storage
- $0.50per GB-month
- Vector storage
- $2.00per GB-month
Overage is postpaid and reconciled at the end of each billing cycle. If the total comes to less than $1, we waive it.
Spend cap
Every paid plan has a spend cap that defaults to twice the plan price — $38 on Starter, $198 on Pro. Reach it and billable reads and writes pause until the cycle resets.
- —Adjust it yourself anywhere from $0 to $500.
- —Set it to $0 for a hard stop at your plan allowance — never pay for overage.
- —Need a higher ceiling? sales@foxnose.net.
On the Free plan
The Free plan has no overage. Each axis stops on its own when its allowance runs out, and the rest keep working — if you exhaust retrievals, writes still go through, and the other way around.
An exhausted axis returns 402 with a machine-readable body your agent can act on, not a rate-limit error. Upgrade to lift the allowance immediately.
Allowances and limits
Usage allowances and structural limits, side by side. Every plan ships the full platform — the difference is scale.
| Free | Starter | Pro | Enterprise | |
|---|---|---|---|---|
| Monthly usage | ||||
| Retrievals | 10,000 | 75,000 | 400,000 | Custom |
| Writes | 5,000 | 20,000 | 100,000 | Custom |
| Data storage | 0.5 GB | 5 GB | 20 GB | Custom |
| Vector storage | 0.1 GB | 1 GB | 4 GB | Custom |
| Overage | Hard stop | Plan rate | Plan rate | Negotiated |
| Structural limits | ||||
| Projects | 1 | 3 | 10 | Custom |
| Environments (per project) | 2 | 3 | 10 | Custom |
| Collections (per environment) | 50 | 200 | 1,000 | Custom |
| Flux APIs | 5 | 10 | 50 | Custom |
| Locales | 2 | 5 | Unlimited | Custom |
| API keys | 5 | 10 | 50 | Custom |
| Custom roles | 0 | 5 | 20 | Custom |
| Rate limit | 120 rpm | 600 rpm | 1,200 rpm | Custom |
| Data & support | ||||
| Data residency | EU | EU | EU | EU |
| Infrastructure | Shared | Shared | Shared | Dedicated database servers |
| Support | Docs | Priority email | SLA | |
Included on every plan
No capability is gated behind a higher tier. Plans differ by scale and production support, not features.
Search & Retrieval
- Hybrid search
- Semantic (vector) search
- Full-text search with typo tolerance
- Automatic embedding generation
- Real-time indexing
- Multi-level filtering
- Cross-resource joins
- Relevance ranking
API & Delivery
- Auto-generated Flux APIs
- Embedded MCP server
- Hierarchical URLs
- Multiple delivery APIs
- Population (nested fetching)
- Localization & fallbacks
Content Modeling
- Visual schema editor
- Content editor
- Markdown editor
- Schema versioning
- Resource versioning
- Validation rules
- Reusable components
- Typed collections
- Parent-child relationships
- Draft & publish workflow
Environments & Governance
- Isolated environments
- Activity log
- Role-based access control
- Scoped API keys
Security & Compliance
- DDoS protection
- GDPR compliance mode
- Request signing
- EU data residency
Developer Experience
- Python & TypeScript SDKs
- Dual API architecture
- LLM & agent integrations
- Dashboard
Frequently asked questions
- What counts as a retrieval and a write?
- A retrieval is one read from a Flux API — a GET, a search, a schema or router call — or one MCP tools/call. It is counted 1:1, so a search returning many results is still a single retrieval. A write happens when you publish content; it triggers embedding generation and is weighted by how much text is embedded. MCP handshake calls (initialize, tools/list, ping) are free.
- Why did one write count as three?
- Writes are weighted by embedded size using writes = max(1, ceil(embedded_tokens / 500)). Content up to roughly 500 embedded tokens is one write; a document with about 1,500 embedded tokens is three. Embedding is the expensive part of a write, so cost tracks the amount of text embedded rather than the request count. Only fields you mark as searchable are embedded.
- How does the spend cap work?
- On a paid plan the spend cap defaults to twice the plan price — $38 on Starter, $198 on Pro. When overage reaches it, billable reads and writes pause until the cycle resets. You can move the cap yourself between $0 and $500; setting it to $0 means a hard stop at your plan allowance with no overage at all. For a higher ceiling, contact sales@foxnose.net.
- What does annual billing save?
- Annual plans are 17% cheaper than paying monthly. Starter is $190 per year (about $15.83/mo) and Pro is $990 per year (about $82.50/mo). You can switch between monthly and annual at any time.
- Can I invite team members?
- Organizations are single-owner today. Role-scoped API keys let you hand out safe, limited access to services and agents, and team member invitations are on the roadmap.
- What does my agent see when an allowance runs out?
- On the Free plan, each axis stops independently and returns HTTP 402 plan_exhausted with the axis name and a reset time — a machine-readable body your agent can handle, not a rate-limit error. The axes that still have allowance keep working. On a paid plan at the spend cap you get 402 spend_cap_reached with the cap amount, the reset time, and a URL to raise it. Over MCP, tools/call returns JSON-RPC error -32003 carrying the same details, while the handshake keeps working.
- What happens to my data if I downgrade or cancel?
- Nothing is deleted. A downgrade only blocks creating resources beyond the new limits — everything already stored stays in place. If you cancel, your account moves to Free at the end of the paid period and your data is retained for at least 90 days.
- What if a payment fails?
- You get a 14-day grace period with full service while the payment is retried. If it still fails after that, the account moves to the Free plan. Your data is never removed during this process.
- When do usage windows reset?
- Every plan, including Free, resets on the same day of the month you signed up or subscribed. Unused allowance does not roll over — each window starts fresh. Storage is sampled roughly every four hours and averaged across the window.
Ready to get started?
Create your free account and start building in minutes. No credit card required — upgrade when your allowance runs out.
Get Started Free