Start free. Then pay for what you actually run.
Ragwalla is a managed AI agent platform: agents, orchestrators, retrieval, knowledge graphs, memory, MCP servers, tool hosting, and seven channels. One metered platform price, no seat minimums, and no markup on your own model keys.
How Ragwalla pricing works
01
Start free. No card.
A real dedicated endpoint, 2,500 execution steps a month, and a $5 AI credit. No credit card, no trial clock, no sales call.
02
Then pay only for what you run.
$2 per 1,000 execution steps, $2 per GB-month of storage, and AI at Ragwalla's published per-model rates. Credits never expire, you set the caps, and a surprise bill is structurally impossible.
03
Bring your own key and we charge nothing for AI.
Connect your own provider key and your model spend goes to that provider at their price. Ragwalla's AI charge is exactly $0.00 — you pay us for the platform, never a markup on your tokens.
04
Everything is included on every plan.
Audit logs, RBAC, backups, retention controls, and BYOK ship with every account, free tier included. We don't paywall the features you need in order to trust us.
05
Subscriptions come later, and only when they save you money.
We will add plans when enough accounts spend enough that a bundle is genuinely cheaper than metered rates. Until then, pay-as-you-go is the entire price list.
Bring your own key and we charge nothing for AI.
Connect your own OpenAI, Anthropic, Google, Azure, Bedrock, Mistral, Groq, or OpenRouter key. Your model spend goes straight to that provider at that provider's price. Ragwalla's AI line item is $0.00 — not a discount, not a lower multiplier, zero.
- AI on your key
- $0.00
- Billed by your provider, never re-billed by us.
- AI on our keys
- Published rates
- Per model, per token, cached tokens included. No hidden multiplier.
- Keys stay yours
- Encrypted
- Envelope-encrypted per project. Revoke at any time.
No credit card
Free tier
$0/month, indefinitely
Not a trial. A real dedicated endpoint with its own database and Durable Objects. When the allowances run out, work stops with a clear error — we can't bill a free account, so a surprise invoice is structurally impossible.
- 1 dedicated endpoint — your own D1 database and Durable Objects, not a shared sandbox
- 2,500 execution steps per month
- $5 one-time AI credit for Ragwalla-key models
- Unlimited BYOK model usage within the step allowance
- 200 MB of storage
- 3 agents, 1 orchestrator, 2 subagents, 1 assistant
- 2 vector stores, 1 memory store, 2 MCP servers, 2 deployed tools
- 1 channel, 1 concurrent run
- 7-day audit retention
Pay-as-you-go rate card
Top up in $10, $25, or $100 blocks. Purchased credit never expires. Auto-recharge is off unless you turn it on, and it obeys a monthly cap you set.
| Meter | Price |
|---|---|
| Execution stepsOne model call, one tool execution, or one delegation. A standalone retrieval API query counts as one step. | $2.00 per 1,000 |
| AI usage — your own key (BYOK)Bring an OpenAI, Anthropic, Google, Azure, Bedrock, Mistral, Groq, or OpenRouter key and we add nothing to your model spend. Platform meters still apply. | $0.00 |
| AI usage — Ragwalla keysOur rate card covers input, output, and cached tokens per model. We reserve the maximum cost of a call before it runs and refund the difference, so overruns are bounded. | Published per-model rates |
| StorageVectors, files, knowledge graphs, and memory stores. Charged daily against measured bytes. 1 GB included on funded accounts, 200 MB on the free tier. Out of credit blocks writes — reads keep working. | $2.00 per GB-month |
| Additional endpointsYour first endpoint is free. Extra endpoints are charged per day, so you only pay for the days they exist. An unfunded endpoint is suspended, never deleted. | ≈ $5.00 per month each |
| Browser and sandbox sessionsPremium execution resources are metered from day one and capped on funded accounts. They are disabled on the free tier rather than silently unlimited. | Metered — off by default on free |
What counts as an execution step?
One execution step is one model call, one tool execution, or one delegation to another agent. A run that calls a model twice, runs three tools, and delegates once is six steps. Internal bookkeeping, retries we caused, and failures on our side are free.
We meter the work, not the conversation. There is no per-conversation charge, no per-seat charge, and no charge that grows just because a thread is long.
We don't paywall trust
Step-level audit trails, RBAC, backups, retention controls, legal holds, and BYOK are included on every account — free tier included. They are never gated behind a higher-priced enterprise tier, because you shouldn't have to pay extra to verify what your agents did.
Also metered at $0
- Memory and knowledge-graph reads
- Webhooks, including delivery retries
- Audit logging and retention policies
- Team seats and RBAC
- Channel connections
- Bandwidth and WebSocket connection time
Where are the monthly plans?
We haven't written them yet, on purpose. A subscription is only worth buying when the bundle costs less than the metered rates, and we don't have the usage data to price one honestly. When we do, plans will show up here with the math next to them — and overage will be charged at exactly the rates above.
Had a subscription before? Nothing to re-buy. Your account moves to pay-as-you-go automatically — add credit whenever you want to keep running, or bring your own model keys and pay us nothing for AI. Open billing.
Need a custom solution?
Contact our sales team to discuss enterprise agreements, custom integrations, dedicated capacity, or specific compliance requirements for your business.