Hot Cache Booster
$49 /mo per 5GB
Reduces replay and read costs.
Detailed pricing
Everything in one place
Full details on platform plans, agent execution, sandbox runtime billing, and Backend-in-a-Box bundles.← Back to simplified pricing
Agent execution
Run tool calls, state, and memory for AI agents on dedicated infrastructure. AI agent capabilities are bundled into all platform plans.
| Plan | Price | AI Calls / Month | Agents | Concurrency |
|---|---|---|---|---|
| Free | $0 | 10K (hard stop) | 3 | 10 req/min |
| Starter | $24/mo | 100K + $0.15/1K overage | 10 | 100 req/min |
| Professional | $79/mo | 1M + $0.08/1K overage | 100 | 500 req/min |
| Enterprise | $299/mo | 5M + $0.05/1K overage | 500 | 2,000 req/min |
| Agent Enterprise | $499/mo | Unlimited | Unlimited | Unlimited |
State Fabric (included in all plans)
| Plan | State Objects | Hot Cache | Multi-Region |
|---|---|---|---|
| Free | 1 (Sandbox) | — | — |
| Starter | 3 | Add-on | Add-on |
| Professional | 10 | Add-on | Add-on |
| Enterprise | Unlimited | Add-on | Add-on |
State Fabric Add-ons
Enhance your State Fabric with premium capabilities. Available on Starter and above.
Multi-Region Replication
$99 /mo
Active-active replication across regions for HA and global latency.
AI Memory Pack
$149 /mo
Vector index, embeddings storage, fast read engine.
Advanced Insights
$79 /mo
Cost forecasting, anomaly detection, hot path alerts.
Advanced Security Pack
$99 /mo
SOC2-friendly logs, key rotation, audit streams.
Add-ons are available on all paid plans (Starter and above). Prices can change—confirm current rates in your billing page.
Always-On AgentsHybrid subscription + per-minute
Run agents around the clock with persistent memory and zero cold starts. LLM costs go directly to your provider via your own API key (BYOK) — the platform fee covers compute, memory, and orchestration only.
| Plan | Included Hours | Active Rate | Idle Rate | Context Pool | Spend Cap |
|---|---|---|---|---|---|
| Free | Not available | — | — | — | — |
| Starter | 50h / month | $0.10 / min | $0.025 / min | 256 MB | $50 / month |
| Professional | 200h / month | $0.08 / min | $0.020 / min | 1 GB | $150 / month |
| Enterprise | 1,000h / month | $0.05 / min | $0.0125 / min | 10 GB | $500 / month |
| Agent Enterprise | Unlimited | $0.03 / min | $0.0075 / min | Unlimited | Unlimited |
How always-on billing works
- Active time is when an agent is processing a request or running a background task — billed at 100% of the per-minute rate.
- Idle time is when an agent is waiting for requests (5+ minutes of inactivity) — billed at 25% of the active rate.
- Context memory pool is shared across all your always-on agents. Exceeding the pool triggers overage charges at the plan's per-GB-hour rate.
- Spend caps pause your agents automatically before runaway usage — you receive an email alert at 80% and a final warning before pause.
- BYOK means your LLM costs go directly to OpenAI, Anthropic, or whichever provider you connect — the platform never marks up inference.
Tip: Mid-month plan upgrades include a prorated credit for unused included hours on your old plan. Downgrades take effect at the start of the next billing cycle.
Sandbox runtime billing
Every sandbox execution pays for its wall time — billed per second against your wallet, with daily per-tenant caps. Includes python-light, node-lite, node22-typescript, typescript-wasm, ruby, FlyBox, and CCP GPU pods. One row per execution in the ledger; per-second resolution is live observability.
| Runtime tier | Per second | Per minute | Per hour (equivalent) | Margin vs COGS |
|---|---|---|---|---|
python-light | $0.0000167 | $0.001 | $0.06 | ~94% |
node-lite | $0.0000333 | $0.002 | $0.12 | ~94% |
node22-typescript | $0.0000167 | $0.001 | $0.06 | ~94% |
typescript-wasm | $0.0000083 | $0.0005 | $0.03 | ~88% |
ruby | $0.0000333 | $0.002 | $0.12 | ~91% |
flybox (s-2vcpu-4gb default) | $0.0003 | $0.018 | $1.08 | ~97% (DO on-demand) |
ccp.l4 (L4 24GB) | $0.0004 | $0.024 | $1.44 | ~70% (RunPod Community) |
ccp.a5000 | $0.0002 | $0.012 | $0.72 | ~78% (RunPod Community) |
ccp.a100 (80GB) | $0.0007 | $0.042 | $2.52 | ~53% (RunPod Community) |
ccp.h100 (80GB) | $0.0012 | $0.072 | $4.32 | ~54% (RunPod Community) |
Daily per-tenant sandbox-second capsRefuse-new admission
When a tenant exceeds their daily cap, new sandbox starts return a typedsandbox_daily_cap_exceeded error. Currently-running boxes always finish their work — the cap is enforced at admission only. The optionalSANDBOX_LOW_BALANCE_KILL_THRESHOLD env var enables hard-kill for ops-driven drain scenarios.
| Plan | Daily cap | Equivalent |
|---|---|---|
| Free | 600 s | 10 min / day |
| Starter | 3,600 s | 1 h / day |
| Professional | 18,000 s | 5 h / day |
| Enterprise | 86,400 s | 24 h / day |
| Agent Enterprise | 86,400 s | 24 h / day |
How sandbox billing works
- Admission — every
StartSpanreads the in-memory per-tenant daily sum and rejects the span withsandbox_daily_cap_exceededif the cap would be breached. - Metering — at execution end, the span writes one row to
sandbox_metering_events(1-second minimum, ceil-rounded to whole seconds). Cardinality is bounded by executions, not seconds. - Rollup — every minute the cron sweeps unbilled events into
cost_allocation_entrieswithkind='sandbox_seconds', then debits the wallet via the existing credits-first / USD-fallback path. - Observability — the
functionfly_sandbox_seconds_buckethistogram carries per-second resolution live; thesandbox_daily_usage_rollupstable feeds the dashboard's Sandbox Usage panel (today's used / cap, in-flight count, EOD projection).
Per-tier pricing is operator-overridable viaSANDBOX_PRICING_<TYPE>_US_PER_SEC env vars or the adminPUT /v1/admin/sandbox/pricing/<tier> endpoint.
Backend-in-a-Box
Skip the decision fatigue. Get pre-configured bundles with Auth, Payments, User DB, and more.Start free until you hit 100 users or $1K MRR.
SaaS Starter — $29/moAuth · Payments · User DB · Email · Analytics
- Pre-configured auth (OAuth, magic links)
- Stripe payments & subscription management
- User DB with role-based access
- Transactional email via SendGrid
- Usage analytics dashboard
Marketplace — $49/moListings · Payments · Messaging · Notifications
- Product listings & search
- Stripe Connect for multi-vendor payments
- Real-time messaging between buyers/sellers
- Push & email notifications
- Order management & dispute resolution
AI App — $39/moVector DB · Embeddings · Chat · Memory
- pgvector for semantic search
- OpenAI & Anthropic embeddings pipeline
- Chat history & conversation memory
- RAG-ready document ingestion
- Token usage analytics
🎯 100 users free💰 $1K MRR free⏰ 3 months free
Ready to ship trusted tools?
Create a workspace, connect providers, and see live limits.