Pricing.
One pricing model per product. Token credits across all.
Pay for what you use. Subscribe to what you keep. Hold a credit balance for model inference. The line below each product spells out what gets billed.
All prices in EUR, VAT excluded. Numbers iterate per product without redeploy; figures publish as each product locks its rate. Page shape stays the same.
One pricing model per product. Token credits across all.
Enterprise Mesh
Consumption.
Pay for what the mesh actually does. Fifteen line items meter the substrate: team compute (vCPU and vRAM), events and orchestration steps + executions, network I/O, team memory and knowledge layers (storage plus read / write ops), and the gateways that proxy tool calls and model inference.
- Team vCPU
- _ € per vCPU-hour
- Team vRAM
- _ € per vRAM-hour
- Events published
- _ € per 100 events
- Message storage
- _ € per GB
- Long-running workflow steps
- _ € per 100 triggered
- Long-running workflow execution
- _ € per execution
- I/O
- _ € per GB, network in and out
- Memory storage
- _ € per GB
- Memory read ops
- _ € per million ops
- Memory write ops
- _ € per million ops
- Knowledge storage
- _ € per GB
- Knowledge read ops
- _ € per million ops
- Knowledge write ops
- _ € per million ops
- Tool gateway
- _ € per million tokens, input and output
- Model gateway
- _ € per million tokens, input and output
Runtime, orchestration, governance, identity, tool gateway, observability, evals, event bus, event store, edge as deployment target.
Book a callWorkSpace
Free.
The work surface itself is free; the work your agents and teams do through it bills as Mesh consumption.
Free. The work bills as Mesh consumption; see the Enterprise Mesh card.
One surface for the whole workforce, people and agents working side by side, TeamStore browsing.
Get startedWorkforce Studio
Consumption.
Mesh consumption, no separate SKU. Building your workforce and running it both bill as the mesh meters them.
- Build
- Mesh consumption while the agents that build your workforce run.
- Run
- Mesh consumption as your workforce works; see the Enterprise Mesh card.
- On top
- Token credits + gateway fee for model inference.
Workforce Studio UI, conversational building, Mesh runtime, governance + audit trail, EU residency.
Book a callLocal Assistant
Fixed monthly per seat.
Fixed monthly per seat. The agent runs locally on your desktop; additional Mesh consumption bills only when it reaches into your company's mesh.
_ € / seat / month, plus Mesh consumption when on-mesh.
Local desktop runtime, Enterprise Mesh integration, on-mesh-when-you-opt-in.
Book a callManaged Hermes Agent
Fixed monthly per agent.
Fixed monthly per agent, all-in. A persistent cloud personal agent hosted by Myrmid in EU-resident infrastructure; one predictable line item covers hosting, management, and security updates. Optional Mesh consumption when on-mesh.
_ € / agent / month, plus Mesh consumption when on-mesh.
Hosted Hermes Agent runtime, persistent capability, Telegram access, management, security updates, EU-resident operation.
Book a callManaged OpenClaw
Fixed monthly per claw.
Fixed monthly per claw. A cloud personal agent hosted off-mesh by Myrmid; one predictable line item covers runtime, management, monitoring, and updates.
_ € / claw / month.
Hosted OpenClaw runtime, monitoring, updates, EU-resident operation.
Book a callToken Credits.
Token credits buy model inference. Each million tokens consumed bills at two parts: the provider's compute cost (no markup from us) plus a gateway fee. The gateway fee is the price of agnosticism and decoupling: it lets you swap models or providers without re-plumbing your teams, and shields you from any single vendor's pricing moves.
Multiple providers, multiple data residencies. Each table below lists the models we offer through that provider and the rate per million tokens, input / output, in EUR, VAT excluded. Provider names link to that provider's data protection terms.
Rates as of 2026-06-08. Scaleway and Mistral API rates are native EUR; AWS Bedrock Frankfurt and Gemini Enterprise Agent Platform Belgium rates are converted from USD source at 0.87 EUR/USD and refresh quarterly.
Gateway fee: _ € per million tokens, applied on input and output, across all providers.
| Model | € / M tokens (input / output) |
|---|---|
| Mistral Medium 3.5 | 1.50 / 7.50 |
| Qwen3.5 397B A17B | 0.60 / 3.60 |
| Devstral 2 123B | 0.40 / 2.00 |
Show all 6 models
| gpt-oss 120B | 0.15 / 0.60 |
|---|---|
| Meta Llama 3.3 70B Instruct | 0.90 |
| Mistral Small 3.2 24B | 0.15 / 0.35 |
| Model | € / M tokens (input / output) |
|---|---|
| Mistral Medium 3.5 | 1.5 / 7.5 |
| Magistral Medium 1.2 | 2 / 5 |
| Mistral Large 3 | 0.5 / 1.5 |
Show all 6 models
| Devstral 2 | 0.4 / 2 |
|---|---|
| Codestral | 0.3 / 0.9 |
| Mistral Small 4 | 0.1 / 0.3 |
| Model | € / M tokens (input / output) |
|---|---|
| Anthropic Claude Opus 4.8 | 4.35 / 21.75 |
| Anthropic Claude Sonnet 4.6 | 2.61 / 13.05 |
| Anthropic Claude Haiku 4.5 | 0.87 / 4.35 |
Show all 6 models
| Amazon Nova Pro | 0.70 / 2.78 |
|---|---|
| Mistral Devstral 2 123B | 0.35 / 1.74 |
| Amazon Nova Lite | 0.05 / 0.21 |
| Model | € / M tokens (input / output) |
|---|---|
| Gemini 2.5 Pro | 1.09 / 8.70 |
| Gemini 3.5 Flash | 1.31 / 7.83 |
| Gemini 2.5 Flash | 0.26 / 2.18 |
Show all 5 models
| Gemini 3.1 Flash-Lite | 0.22 / 1.31 |
|---|---|
| Gemini 2.5 Flash-Lite | 0.09 / 0.35 |
Talk to us.
| Enterprise pilot, volume commit, custom SLAs? | Talk to sales |
|---|---|
| Design Partner Program, five partners, working relationship? | Apply |
| Non-profit, public interest? Manifesto Article XIV, we pass our cost through. | Apply |
FAQ.
How does consumption pricing work?
You pay for what the mesh actually does: team compute (vCPU and vRAM), events and orchestration, network I/O, memory and knowledge, and the gateway calls. Billed monthly. Ranges and unit pricing land before GA; today we commit to the model structure.
What are token credits?
Token credits buy model inference. Each million tokens consumed bills as the provider's compute cost (no markup from us) plus a gateway fee. The fee is the price of agnosticism: swap providers without re-plumbing your teams.
Why is WorkSpace free?
WorkSpace is the surface where your people and agents work; the real cost is the work that runs through it, billed via Mesh consumption. Pay for the work, not the surface.
Can I commit annually for better rates?
Yes. Annual contracts include volume commitments and locked rates. Enterprise contracts also include custom SLAs and onboarding support. Contact us for terms.
What's the difference between Managed OpenClaw and Local Assistant pricing?
Managed OpenClaw is fixed monthly per claw; Myrmid hosts and runs the agent for you (off-mesh by design). Local Assistant is fixed monthly per seat plus Mesh consumption when your local agent reaches into the company's mesh.
Currency and VAT?
All prices in EUR, VAT excluded. Quoted figures are net; regional VAT is applied at invoicing per EU and EEA rules. Non-EU customers contact us for invoicing terms.