Pricing
Predictable pricing that scales
Start free, bring your team, and scale to millions of runs.
Free
$0always
Get started with durable execution and full observability, free, no credit card.
Get started for freeIncluded usage
- 250K tasks / mo
- 50 concurrent runs
- 3 connected workers
- Unlimited users
Features
- Logs, traces, full run inspection
- LLM token cost attribution
- 7-day run history
- Unlimited projects
- Community support
Pro
[ Most popular ]$99starting at / mo
Everything you need to run AI agents and workflows in production with confidence.
Get started for freeIncluded usage
- 5M tasks / mo
- 1,000 concurrent runs
- 25 connected workers
- Unlimited users
Features
- Everything in Free
- Evals & human-in-the-loop approvals
- Custom dashboards and alerting
- 90-day run history
- Email support
Enterprise
Contact usto learn more
For teams that need scale, control, and a direct line to our engineering team.
Request a demoIncluded usage
- Custom volume
- Custom concurrency
- Custom workers
- Unlimited users
Features
- Everything in Pro
- SSO / SAML
- Audit logs
- Dedicated infra
- Custom retention
- Data residency
- SLA + dedicated Slack channel
Compare plans
Every feature, by plan
| Feature | Free | Pro | Enterprise |
|---|---|---|---|
| Core | |||
| Base price | $0 | $99 / mo | Contact us |
| TasksEach step in your code: function call, LLM call, tool use, signal, or wait | 250K / mo included | 5M / mo included, then volume-tiered | Custom |
| Concurrent runsParallel executions in flight at any moment | 50 concurrent | 1,000 concurrent | Custom |
| Connected workersSDK processes holding live connections | 3 workers | 25 workers | Custom |
| Gateway throughputAPI requests per second | 25 RPS | 500 RPS | Custom |
| Webhooks | 3 | Unlimited | Unlimited |
| Crons | 3 | Unlimited | Unlimited |
| Environments per project | 1 | Unlimited | Unlimited |
| Projects, components, service keys | Unlimited | Unlimited | Unlimited |
| Evaluations & Human-in-the-Loop | |||
| Eval runsBilled as tasks | Included | Included | Included |
| Custom scorersUser-defined evaluation functions | 1 scorer | Unlimited | Unlimited |
| Eval datasets | Unlimited | Unlimited | Unlimited |
| Human approvals (HITL)Pause and resume runs for human review | Basic | Full | Full |
| Run annotations & sharing | Basic | Full | Full |
| Comparison & A/B run views | Not included | Included | Included |
| Observability | |||
| Logs, traces, full run inspection | Included | Included | Included |
| LLM token attributionCost by model, by component, by run | Included | Included | Included |
| Cost & latency dashboards | Included | Included | Included |
| Metrics granularityReal-time aggregation window | 5 minutes | 1 minute | 1 minute |
| Run & trace historyTime you can look back at every run | 7 days | 90 days | Custom |
| Custom dashboards & alerting | Not included | Included | Included |
| Trace exportPush to Datadog, Honeycomb, OpenTelemetry | Not included | Add-on | Included |
| Security & Support | |||
| Tenant isolationPer-tenant key prefix in storage; row-level filter in control plane | Included | Included | Included |
| TLS in transit, encryption at rest | Included | Included | Included |
| Users | Unlimited | Unlimited | Unlimited |
| API keys & service keys | Unlimited | Unlimited | Unlimited |
| SSO / SAML | Not included | Not included | Included |
| Audit logsTrack every change to projects, deployments, keys | Not included | Not included | Included |
| Dedicated infrastructureSingle-tenant data plane | Not included | Not included | Included |
| Data residencyChoose region for storage and compute | Not included | Not included | Included |
| Support | Community | SLA + Dedicated Slack | |
| Onboarding & migration help | Self-serve docs | Self-serve docs | Hands-on |
| Uptime SLA | Not included | Not included | 99.9% |
FAQ
Frequently asked questions
What's a task?
A task is each step your code performs — a function call, an LLM call, a tool invocation, a signal received, or a timer firing. A simple workflow uses a handful of tasks. A complex agent run uses dozens. You see the count in your dashboard for every run.
Why charge per task instead of per workflow?
Because that's what we actually run. A workflow with 5 steps does 5× the work of one with 1 step — billing the same for both would either overcharge simple cases or undercharge heavy ones. Per-task pricing is honest, and the tiered volume discount means heavy users pay less per unit as they grow.
Do I pay for LLM tokens?
No. You pay your model provider (OpenAI, Anthropic, etc.) directly. AGNT5 includes full LLM token observability — cost-by-model, cost-by-component, latency breakdowns — at every tier, including Free. The tokens are yours; the visibility is ours.
What happens if I exceed my plan?
On Pro, tasks beyond the included 5M are billed at the volume-tiered rates above. No throttling, no surprise outages — your code keeps running. You'll see usage projections in your dashboard and an alert at 80% of your bundle so there are no surprises.
Can I start on Pro without paying upfront?
Yes. Click “Get started for free” on Pro and you'll have full access to all Pro features under the Free allotment. You only enter a card when your usage exceeds Free limits — at that point you're upgraded to Pro automatically.
How do I switch between plans?
Free → Pro is automatic when you exceed the Free tier's task limit (with your consent). Pro → Enterprise is a conversation with our team — typically when you need SSO, dedicated infrastructure, custom retention, or volumes above 200M tasks / month.
Is there a free trial of Enterprise features?
Enterprise features (SSO, audit logs, dedicated infrastructure, custom retention) are tied to deployment changes that require coordination with our team. Reach out and we'll set up a proof-of-concept on a dedicated environment.
Do you charge per user or per seat?
No. Users are unlimited on every tier, including Free. Your bill scales with what your code actually does, not how many people are looking at it.
What's “concurrent runs” and when would I hit the limit?
A run is one execution of a workflow, agent, or function. “Concurrent” is how many can be in flight at the same time. Free's 50 covers most prototypes. Pro's 1,000 covers nearly all production workloads. If you're consistently bumping the ceiling, that's an Enterprise conversation.
What happens to my data if I cancel?
You can export all your runs, traces, and metrics for 30 days after cancellation. After that, data is permanently deleted from our systems on the schedule defined by your plan's retention policy.
Do you offer annual discounts?
Yes — annual prepay gets a 20% discount on the base fee. Email billing@agnt5.com to arrange.
How is my data isolated from other customers?
Every tenant has a unique key prefix in our durable storage and a unique row-level filter in our control plane database. There is no cross-tenant query path. On Enterprise, you can opt for fully dedicated infrastructure with no shared data plane at all.
Questions? Reach out: sales@agnt5.com
Deploy your first agent in 5 minutes.
Start free with 250K steps a month.