PricingEarly access, September 2026
Every plan runs the whole platform: workflows, agents, approvals and the audit trail. Plans differ in how much you can run. During early access we set the price with you for each pilot, and there’s no self-serve sign-up yet.
A fourteen-day evaluation on any plan, set up with the team. Upgrade when you outgrow it. Talk to sales when you need a custom contract.
Plan tier is a capacity band on the tenant (TenantPlan: STARTER, TEAM, PROFESSIONAL, ENTERPRISE), not a feature gate. Paid modules will use a separate entitlement layer, designed and not yet built.
Try it: send a run to a person and watch its place free up.Toggle PAY-2291 to WAITING_FOR_APPROVAL.
01Plans
Every plan runs workflows, agents, approvals and the audit trail. Higher plans add capacity, retention, support and controls for larger teams.
For small teams getting started with workflow automation.
What’s included
Best for teams under 5
For teams standardising automation across departments.
Everything in Starter, plus
Best for 5–15 people
For growing teams running production workflows.
Everything in Team, plus
Best for production teams
For regulated industries and large-scale deployments.
Everything in Professional, plus
Best for regulated industries
Every plan includes
The basics that aren’t negotiable.
If a capability isn’t listed, it’s on the roadmap or available as an add-on. Ask and we’ll tell you which.
| Capability | Starter | Team | Professional | Enterprise |
|---|---|---|---|---|
| Authoring | ||||
| Visual canvas | Included | Included | Included | Included |
| JSON definitions | Included | Included | Included | Included |
| Python SDKTo confirm | Included | Included | Included | Included |
| Custom activities | – | Included | Included | Included |
| Agents | ||||
| agent.run | Included | Included | Included | Included |
| MCP inbound (expose workflows) | – | Included | Included | Included |
| MCP outbound (external tools) | – | – | Included | Included |
| Sub-workflows as tools | – | – | Included | Included |
| Humans | ||||
| Approval gates at any step | Included | Included | Included | Included |
| Slack, Teams and email | Included | Included | Included | Included |
| Magic-link approvals | Included | Included | Included | Included |
| Escalation and quorum policiesTo confirm | – | Included | Included | Included |
| Security | ||||
| MFA and step-up auth | Included | Included | Included | Included |
| SSO (SAML / OIDC) | – | Included | Included | Included |
| SCIM provisioningNot offered | – | – | – | Included |
| IP allowlistNot offered | – | – | Included | Included |
| Data residency choiceNot offered | – | – | – | Included |
| Operations | ||||
| Monitoring and Prometheus | Included | Included | Included | Included |
| Audit log retention | 90 days | 180 days | 1 year | 7 years |
| SIEM streaming exportNot offered | – | – | Included | Included |
| SLA | – | 99.5% | 99.9% | 99.99% |
How usage is counted
A job is one workflow run, however many steps it has. Sub-workflows count toward their parent, and an agent run counts as one job.
One job per run; agent.run is one job regardless of turns or tool calls.
How many runs can be working at the same moment. A run waiting for a person doesn’t take up a place.
Per-tenant max_concurrent_jobs; WAITING_FOR_APPROVAL holds no slot (F1a).
A seat is one person with workspace access. API keys, service accounts and webhook senders don’t use seats.
Seats count human principals only.
During early access, the price is set with you for each pilot.
To confirm marks items still being checked against the product.
Contact sales about Enterprise02Estimate
Set how much you run. The estimate picks the smallest plan that fits and shows what it costs.
Runs waiting for a person don’t count towards concurrent runs. Model costs pass through at the provider rate and aren’t part of this estimate.
Fig. 2 Plan fit for the usage aboveSeat add-ons and agent model spend not included
0203Usage
You shouldn’t need to wait for an invoice to know where you stand. The usage page shows your plan and what you’ve used this period, and agent runs show their cost down to the turn.
max_cost_usd, one of four budgets on every agent run.0304Questions
Not yet. Lyhnis runs as a scoped pilot while the managed cloud is brought live. Book a walkthrough and we’ll set one up with you.
Because we’re setting them with our first customers. We’d rather show no number than one we’ll change next month.
A job is a single workflow run, regardless of how many activities it contains. A workflow with 50 steps is still one job. Sub-workflows invoked from a parent count toward the parent.
An agent run counts as one job no matter how many turns or tool calls it makes. Model costs pass through at the provider rate with no markup; you bring your own key.
A seat is one person with workspace access. API keys, service accounts and webhook senders don’t use seats.
Three factors: job volume, deployment model and support level. Most contracts start with a discovery call.
Yes, for early-stage startups under $1M ARR, registered non-profits and accredited institutions. Discounts run from 30% to 75% depending on the programme. Email sales.
Yes. Every agent run records its tokens and cost per turn, and you can give each agent a cost limit it can’t go past.
Yes. Modules for specific jobs will be sold separately from your plan, so you only pay for the ones you use. The first is in design now.
0405Status
Live means in the product today. Anything still being built is labelled as such, and anything we don’t offer is listed so you can ask before you buy.
Live5In development1Roadmap4Not offered1
As of September 2026
Nothing on the roadmap is claimed anywhere on this site.
If you need one of these now, ask us before you buy.
30-minute call