Start free with a hard cap. Upgrade for scale.
$49/month protects a live application — 50K requests, every alert channel, and project + model attribution. Upgrade when ‘alert me later’ is no longer enough.
One uncapped loop can cost more than a year of Solwyn. You only need the cap to work once.
Ten launch partners get founder-led onboarding, direct implementation support, and a 12-month price lock.
One hard spending cap for your first agents.
- ✓5K requests/month
- ✓3 projects
- ✓1 hard spending cap
- ✓Cost attribution by project
- ✓7-day cost history
- ✓Email alerts
No credit card. No expiration. No catch.
Protect a live AI application.
- ✓50K requests/month
- ✓10 projects
- ✓Hard caps on every project
- ✓2 team members
- ✓Cost attribution by project + model
- ✓30-day cost history
- ✓All alert channels
Cancel anytime. 30-day money-back guarantee.
Shared visibility and per-agent attribution across your team.
- ✓500K requests/month
- ✓Unlimited projects
- ✓Hard caps on every project
- ✓Automatic failover
- ✓5 team members
- ✓Cost attribution + team breakdown
- ✓90-day cost history
- ✓All alert channels
- ✓30-day audit log
- ✓Priority email support
Add agents anytime. No per-seat surprises.
Beyond Team, or need SSO, VPC deployment, custom retention, or contractual support terms? Talk to the founder.
One uncapped retry job burned 32M tokens in 24 hours. A daily cap ends that run at your limit.
Prompt content is never sent to Solwyn — only usage metadata reaches us. Read the open-source SDK →
WHAT HAPPENS WHEN YOU EXCEED YOUR PLAN?
Overage is soft — your agents never stop because of billing. Charges accrue and appear on your next invoice.
Free has no overage billing: past 5K requests, calls pass straight through — unmetered and unprotected — until you upgrade. Your agents are never blocked.
Two different limits: your project budget is the dollar cap you set on an agent's model spend. Your request allowance is what your Solwyn plan includes. Exceeding a paid allowance never turns off budget protection — overage just accrues to your next invoice.
What counts as a request: one LLM-provider call evaluated by the Solwyn SDK before it is sent. Streaming counts as one request. It is not a token, a trace, or an agent step.
Scroll to compare tiers →
| FREE | PRODUCTION | TEAM | |
|---|---|---|---|
| CORE | |||
| Requests / month | 5K | 50K | 500K |
| Projects | 3 | 10 | Unlimited |
| Team members | 1 | 2 | 5 |
| Hard spending caps | 1 cap | Every project | Every project |
| Automatic failover | — | — | ✓ |
| Cost history | 7 days | 30 days | 90 days |
| ALERTS | |||
| ✓ | ✓ | ✓ | |
| SMS | — | ✓ | ✓ |
| Slack | — | ✓ | ✓ |
| Microsoft Teams | — | ✓ | ✓ |
| Telegram | — | ✓ | ✓ |
| Discord | — | ✓ | ✓ |
| PagerDuty | — | ✓ | ✓ |
| Webhook | — | ✓ | ✓ |
| SUPPORT | |||
| Support | Community | Priority email | |
| PRIVACY | |||
| Prompt content never sent to Solwyn | ✓ | ✓ | ✓ |
How the cap behaves
Alerts by default. Hard stops when you want them.
Alerts only — Solwyn never breaks your agents. You get notified; calls keep flowing.
Hard deny — over-budget calls stop in your process, before they reach the provider.
Stays denying — a Solwyn outage won't lift a hard cap that's already in effect.
Read exactly how budget enforcement behaves →
Solwyn caps LLM calls made through the SDK. It is not a secret-scanning tool, a stolen-key firewall, or a cloud-infra waste platform.
Questions & answers.
A request is one LLM-provider call evaluated by the Solwyn SDK before it is sent. Streaming counts as one request. It is not a token, a trace, or an agent step.
Solwyn wraps any LLM client that follows standard API patterns. OpenAI, Anthropic, Google, Mistral, Cohere, and any OpenAI-compatible provider work out of the box. If you can call it from Python, Solwyn can wrap it.
Today, yes — the SDK is Python, installed with pip. More languages are on the roadmap.
Your agents keep running — Solwyn never interrupts service over billing. On paid plans, overage accrues automatically and appears on your next invoice. On Free, there's no overage billing: past 5K requests, calls pass straight through — unmetered and unprotected — until you upgrade.
Never. This is architecture, not a policy promise. The SDK wraps your client locally — LLM calls go directly from your app to the provider, and Solwyn is never in the request path. Only usage metadata (token counts, costs, model, latency, status) reaches us. Prompt and response content is not sent to Solwyn — read the open-source SDK to verify exactly what leaves your process.
No network hop. The budget check happens in your process against cached limits, and your call goes straight to the provider — Solwyn is never in the request path.
Your agents keep running. The SDK enforces cached budget limits locally and continues allowing requests. We fail open by default — Solwyn going down should never take your agents down with it. The one thing an outage never does is lift a hard cap that's already denying. Full details: solwyn.ai/docs/budget-enforcement-semantics.
Start with the free tier — 5K requests/month with full protection, no credit card required. If you outgrow it, upgrade. And because Solwyn is a thin SDK wrapper, removing it takes one line of code — no migration, no lock-in, no cleanup.
Yes. Upgrades take effect immediately. Downgrades take effect at the start of the next billing cycle.
Yes. Annual billing saves you two months — Production drops to $41/mo and Team to $124/mo. Use the toggle on the pricing page to see annual rates.
Ship agents with brakes, not crossed fingers.
Wrap your client. Set a cap. When a loop, retry storm, or job crosses the limit, Solwyn denies the next request before it reaches the provider.
Add a hard cap free30-day money-back guarantee on all paid plans. Refund policy