Relay Pricing
One published rate card for every infrastructure. Usage, complexity and service are priced separately — so you can budget a resilient architecture without a sales call.
Three things carry the price
Nothing else does. Where Relay runs never changes what it costs.
Usage
Packets moved and node-hours held. Billed per minute, only while a node is running.
Complexity
Which reliability capabilities you switch on — cluster mode, federation, DR, durability, delivery guarantees. Published per module.
Service
Setup as a one-time engagement, maintainability as a subscription. You operate the fleet; we keep it upgradeable and healthy.
Plans
Every plan runs on AWS, Azure, GCP, your own racks, or an air-gapped enclave — at the same rates.
Free
Prove high availability on your own hardware
- 25M packets / month
- 2 nodes · 1 cluster
- Cluster mode included — real failover, not a single broker
- stantia-cli and dash included
- Community support, no SLA
- Never hard-stops your message flow
Scale
Production, metered, no minimum
- usage
- + capabilities
- + service plan
- Unlimited packets, nodes, clusters, connections
- All capability modules à la carte
- Same rates on any infrastructure
- Standard maintainability included
- No platform fee, no seat fee, no minimum
- DPA, BAA, sub-processor notice, audit rights
Sovereign
Regulated, air-gapped, on your own terms
- same usage
- + same capabilities
- + Mission Critical
- Identical usage and capability rates to Scale
- Air-gap capable — offline licence and metering
- FIPS mode, STIG/CIS baseline, signed SBOM
- Validated hardware and kernel matrix
- Source escrow, exit plan, regulator access
- Named TAM, 24×7, monthly health assessment
Axis 1 — Usage
Packets moved, node-hours held. Identical on every infrastructure target.
Packets
One packet is one message accepted at ingress, up to 4 KiB of payload. Egress fan-out is included up to 20 deliveries per ingress packet.
| Monthly volume | Per million packets |
|---|---|
| First 100M | $0.30 |
| 100M – 1B | $0.18 |
| 1B – 10B | $0.09 |
| 10B – 100B | $0.045 |
| Above 100B | $0.022 |
| Excess fan-out beyond 20:1 | $0.015 |
Node-hours
Billed per minute from first accepted connection to process exit, with a five-minute minimum. Scale down at 02:00 and you stop paying at 02:00.
| Class | Capacity envelope | Per node-hour | ≈ per node-month |
|---|---|---|---|
| S | ≤ 4 vCPU · ≤ 16 GiB | $0.040 | $29 |
| M | ≤ 16 vCPU · ≤ 64 GiB | $0.160 | $117 |
| L | ≤ 48 vCPU · ≤ 192 GiB | $0.470 | $343 |
| XL | ≤ 96 vCPU · ≤ 384 GiB | $0.870 | $635 |
| XXL | > 96 vCPU | $0.009 / vCPU-hr | — |
| Passive / standby | Any class, no client traffic | 25% of class | |
| Dev / test / staging | Up to 25% of production node count | $0.00 | |
Never billed
Connections, channels, topics, subscriptions, retained-message redelivery, intra-cluster replication, storage, egress, keepalives, QoS acknowledgements, health probes, rejected messages, and any traffic duplicated by a Relay defect.
Axis 2 — Complexity
Reliability capabilities, priced per module per production cluster. Switch them on and off from stantia-cli or dash.
| Module | What it turns on | Per production cluster / month |
|---|---|---|
| Cluster | Multi-node clustering, shared session state, automatic in-site failover, rolling upgrade without message loss | $240 |
| Federation | Cross-region and cross-site topology, geo-aware routing, inter-cluster bridging, topic namespace federation | $420 |
| Disaster Recovery | Cross-site replication, orchestrated failover and failback, RPO/RTO targets, DR drill tooling. Includes the paired standby cluster at no charge. | $560 |
| Durability & Replay | Persistent message store, time-travel replay, tiered storage offload, per-topic retention policy | $380 |
| Delivery Guarantees | Exactly-once delivery, strict ordering across partitions, transactional publish, idempotent producers | $310 |
| All five | Full reliability stack on one production cluster | $1,910 |
No base cluster fee
A group of nodes with no capabilities enabled costs nothing beyond its usage. You pay for resilience machinery you actually switched on.
Resilience is discounted, never penalised
DR standby clusters carry no separate capability fee. Passive nodes bill at 25%. Non-production clusters may enable any capability at 25% of list, so failover drills happen in a realistic environment.
Axis 3 — Service
Setup is a one-time engagement. Maintainability is a subscription. They are priced separately because they are different things.
Setup and implementation
| Engagement | What it covers | One-time |
|---|---|---|
| Self-serve | Docs, reference topologies, stantia-cli quickstarts, community | $0 |
| Guided Launch | Architecture review, sizing and class selection, topology design, CLI-driven build of your first production cluster, operator handover, runbook | $12,000 |
| Migration Program | Guided Launch plus incumbent topic and schema mapping, protocol-compatibility assessment, dual-run design, cutover plan and rehearsal, rollback procedure | $35,000Credited against year one |
| Sovereign Deployment | Migration Program plus bare-metal and kernel validation, air-gap installation, FIPS enablement, STIG application, escrow and change-board documentation | $60,000 |
| Additional engineering | Custom platform validation, performance tuning, bespoke integration | $2,400 / day |
Maintainability plans
| Plan | One-time | What you get | Sev-1 response |
|---|---|---|---|
| Community | $0 | Forums, docs, security advisories | — |
| Standardincluded with Scale | $0 | 24×5 ticket support, unlimited contacts, 12-month version support window, upgrade notes | 4 h |
| Business | 12%of metered spend · min $1,500 | 24×7, phone escalation, quarterly capacity and config review, upgrade planning, 24-month LTS window, DR drill review | 30 min |
| Mission Criticalincluded with Sovereign | 22%of metered spend · min $8,000 | Business plus named TAM, escalation executive, monthly health assessment and runbook maintenance, 36-month LTS window, air-gap support logistics | 15 min |
For a platform you run yourself, the question is not whether someone answers a ticket — it is whether you can still upgrade this in three years, and whether someone tells you before you run out of headroom. That is what these plans are for.
Protocols and transports
Every protocol and every transport is available in every plan, including Free. Relay never paywalls a protocol.
Transport layer
| Transport | Implementation | Status | Specification |
|---|---|---|---|
| TCP | Raw stream handling | Implemented | — |
| TLS 1.3 | Full handshake, record layer, cipher suites, certificates, OCSP and CRL revocation | Implemented | RFC 8446 |
| DTLS 1.3 | Anti-replay, handshake and record layer, sharing the TLS crypto core | Implemented | RFC 9147 |
| WebSocket | HTTP/1.1 upgrade, frame encoding and decoding, ping/pong, close handshake | Implemented | RFC 6455 |
| WSS | WebSocket layered over the in-house TLS 1.3 stack | Implemented | RFC 6455 + RFC 8446 |
| QUIC | Streams, flow control, ALPN, datagrams and multipath, via Cloudflare quiche | Integrated | RFC 9000 / 9001 / 9002 |
| CoAP / DTLS | CoAP carried over a DTLS transport. Application payloads are not yet encrypted — do not use where confidentiality is required. | Partial | RFC 7252 + RFC 9147 |
| UDP | Datagram I/O | Functional | — |
| XDP | AF_XDP kernel bypass with eBPF, feature-gated | Implemented | — |
Application protocols — IoT and messaging
| Protocol | Implementation | Status | Specification |
|---|---|---|---|
| MQTT v3.1.1 / v5.0 | Full packet parser, session store, QoS 0/1/2, authentication and authorization | Implemented | OASIS MQTT v3.1.1 / v5.0 |
| CoAP | Message parser, options, confirmable and non-confirmable messages | Implemented | RFC 7252 |
| MQTT-SN | Packet parser and gateway mode | Implemented | OASIS MQTT-SN v1.2 |
| DNS | Message parser with question, answer, authority and additional sections | Implemented | RFC 1035 |
| Modbus | TCP frame parser and function codes | Implemented | Modbus App Protocol v1.1b3 |
| SNMP | BER encoder and decoder, message parser | Implemented | RFC 1157 / RFC 3416 |
| STOMP | Frame parser and command handling | Implemented | STOMP 1.0 / 1.1 / 1.2 |
| AMQP | Frame parser, methods and content bodies | Implemented | AMQP 0-9-1 |
| NATS | Message parser and protocol commands | Implemented | NATS Protocol |
| gRPC | Length-prefixed message framing | Implemented | gRPC over HTTP/2 |
| WebSocket | Frame-level protocol handling | Implemented | RFC 6455 |
| OPC UA | Binary message parser, hello/acknowledge/error, services | Implemented | OPC UA Part 6 |
| DDS / RTPS | RTPS message and submessage parser | Implemented | OMG DDS-RTPS v2.5 |
| HTTP/2 | Frame parser with HPACK header compression | Implemented | RFC 9113 / RFC 7541 |
| HTTP/3 | Frame decoder, QPACK header compression, varint encoding | Implemented | RFC 9114 / RFC 9204 |
| WebRTC | SCTP and DCEP DataChannel parser | Implemented | RFC 8832 / RFC 9260 |
Application protocols — AI agents
| Protocol | Implementation | Status | Specification |
|---|---|---|---|
| JSON-RPC 2.0 | Hand-rolled encoder and decoder, newline-delimited and content-length framing | Implemented | JSON-RPC 2.0 |
| MCP | Model Context Protocol — 15 methods, session state machine, SSE streaming | Implemented | MCP 2025-03-26 |
| A2A | Google Agent-to-Agent — AgentCard, task lifecycle, 7 methods, SSE | Implemented | Google A2A |
| ACP | Agent Communication Protocol — agent descriptors and run lifecycle | Implemented | ACP Spec |
| Agent Mesh | Relay-native binary protocol — sub-millisecond agent coordination and registry | Implemented | Relay Internal |
Datapath and acceleration
How the transports are actually served. Available on every plan; the supported hardware matrix determines which acceleration paths are validated for your nodes.
| Layer | What it does | Availability |
|---|---|---|
| XDP / AF_XDP | Kernel-bypass packet processing at the driver layer, before the network stack | All plans |
| io_uring | Thread-per-core asynchronous I/O with submission and completion queues | All plans |
| Katran-style L4 | XDP-based load balancing and connection distribution across nodes | All plans |
| QUIC | Stream multiplexing without head-of-line blocking, connection migration across networks | All plans |
| TLS 1.3 / mTLS | Mutual authentication, session resumption, hardware offload where the NIC supports it | All plans |
Same rate card, any infrastructure
Relay is software you deploy and operate with stantia-cli. Stantia provides the implementation and the deployment path; nothing runs inside your network on our behalf, and no target costs more than any other.
- AWS
- Azure
- Google Cloud
- Your own racks
- Edge sites
- Air-gapped
Included in every plan
- Every protocol and transport, with no paywall
- mTLS, RBAC and SSO/SAML/OIDC
- Tamper-evident audit log
- stantia-cli and the dash control plane
- At-least-once delivery
- Customer-managed keys — BYOK and HYOK
- Published exit procedure and data portability
- Usage records you can verify yourself
Billing and commitments
- Per-minute node metering, five-minute minimum
- No platform fee and no minimum on Scale
- Committed use: 15% at 12 months, 25% at 24, 32% at 36
- Budget caps and hard ceilings on nodes, clusters and packets
- Capabilities enabled during a declared incident are never billed
- Non-payment never stops your data plane or disables a capability
- PO-referenced invoicing, net 45/60, multi-currency
- Rates never vary by infrastructure target
Switching from something else
You are not comparing our price to your bill. You are comparing it to the risk of moving. So we take that risk off the table.
90-day dual run, free
Run Relay alongside your current platform for 90 days with all usage and capability charges waived. Nobody cuts over a production bus on a weekend.
Migration credited back
The Migration Program fee is credited against your first year of usage and capability charges.
Contract overlap credit
Locked into remaining term elsewhere? We credit up to 6 months of overlapping spend, to $150,000.
Published exit
No termination penalty, no egress fee, documented data portability, CLI export of topology and configuration in open formats.
Rates locked for 24 months
From your first production packet, for anyone arriving through the Migration Program. The number you built your case on is the number you keep.
Frequently asked questions
Does it cost more to run Relay on our own hardware?
No. The rate card is identical on AWS, Azure, GCP, your own racks and in an air-gapped enclave. Moving between them carries your rates unchanged, and we commit to that contractually.
Do you operate or monitor our servers?
No, and we hold no standing credentials into your environment. We provide Relay itself, stantia-cli, reference topologies, the validated platform matrix, and support. You deploy, scale and monitor. Diagnostic access is break-glass: you initiate it, it is time-boxed, and it is revoked automatically.
What happens if we exceed something?
There are no overage fees, because nothing is capped on Scale. Set budget caps and hard ceilings on node count, cluster count and monthly packets, and dash shows projected spend before you commit a change. Committed accounts can also burst to 2× for 72 hours a month at no charge.
Is high availability really included in the free plan?
Yes. Free runs a real two-node cluster with automatic failover, so you can prove resilience on your own hardware before you talk to anyone. Most platforms make you buy an enterprise plan before you can test that.
Do you offer an uptime SLA?
Not on the data plane, because you operate it — we would be promising uptime for servers we cannot see or touch. We commit to 99.9% control-plane availability, support response and restoration targets by severity, security patch turnaround, and defect resolution. Where we have reviewed and signed off your topology, an availability warranty is available.
What if the licence expires or we lose connectivity?
Nothing stops. On Scale and Sovereign, licence expiry, clock skew or total loss of control-plane connectivity never halts message flow and never disables an enabled capability. The node logs it, raises an alarm and shows a banner in dash. This is contractual, not a courtesy.
Price your architecture before you talk to us
Every number on this page is published. Build your budget case first — that is rather the point.
All prices in USD, excluding tax. Rates are list prices; committed-use discounts apply from 12 months. Infrastructure you provide is billed by your own provider and never by Stantia Labs.