Your agent's spending department

Flat price. No metered surprises.

Hosted Team gives every agent one deterministic control point, unlimited evaluations, and an API key immediately after checkout.

Team · recommended
$99 / month
  • Unlimited transaction evaluations
  • All rule types + human approval queue
  • Dashboard, audit log, MCP, HTTP + CLI
  • API key issued immediately after payment
Start Team — $99/mo

🛡️ Green-light an active-rule violation and that month is free

Playground
$0

Run sample transactions through the live production firewall. No signup, install, or card.

Try the live firewall →
Business
$499 / month
  • Everything in Team
  • Managed policy onboarding
  • Policy review and rule setup
  • Priority email support
Start Business →

🛡️ Guarantee: green-light a rule violation, month is free

Prefer your own infrastructure? Self-host the MIT-licensed core free forever.

What you're paying for, line by line

No per-call fees. No usage tiers. No surprise invoices when your agent fleet grows.

CapabilitySelf-host (free)Team $99/moBusiness $499/mo
Transaction evaluationsUnlimitedUnlimitedUnlimited
Rule types (per-tx, daily, velocity, merchant, category, time)✓ All✓ All✓ All
Human-in-the-loop approval queue
Queryable SQLite audit trail✓ Local✓ Hosted✓ Hosted
MCP tool, HTTP API, CLI
Dashboard✓ Local✓ Hosted✓ Hosted + priority support
Managed spend policy (we write your rules)
Rule-integrity guarantee✓ Month free✓ Month free

Pricing FAQ

· Last updated 2026-07-17

Is there a free plan?

Yes. The entire open-source core is MIT-licensed and free to self-host forever — full policy engine, dashboard, audit trail, MCP tool, HTTP API, and CLI. The hosted plans add managed infrastructure, persistent log retention, and support.

What counts as a "transaction evaluation"?

One call to the /v1/transactions/evaluate endpoint — your agent asking "should I spend $X at merchant Y?" Each plan includes unlimited evaluations. There is no per-call fee, no metering, no overage charge.

What is the rule-integrity guarantee?

If sipi.bot returns "approve" for a transaction that violated one of your active rules, that month's subscription is refunded. The guarantee covers rule-evaluation engine errors, not losses from rules you didn't configure or from acting against a block/flag decision.

What happens immediately after payment?

Stripe returns you to a private sipi.bot activation page where your API key is displayed with a copy button, a first protected transaction, and framework-specific integration links.

How do plan changes and cancellation work?

Email sales@sipiteno.com from the billing email. We will confirm the effective date before the next renewal; there is no long-term contract.

What if a legitimate transaction is blocked?

Every block includes the exact rule and reason. Adjust or disable that rule in the dashboard, then retry the transaction. A blocked evaluation never moves money itself.

How fast is a transaction evaluation?

The policy check is deterministic local code with no model inference. End-to-end hosted latency depends on your network path and current service load, so measure it from your deployment region before putting it on a critical payment path.

Why trust sipi.bot with your agent's spending

Deterministic, not probabilistic

The rules engine is pure logic — no ML guessing. If a rule says block at $500, every $501 transaction is blocked, every time, with a reason you can audit.

Open source, auditable core

The exact code running the hosted service is on GitHub (MIT). You can read every rule-evaluation path and self-host the same engine.

Queryable audit trail

Every evaluation records its decision, rule, reason, amount, merchant, agent identity, and timestamp in SQLite for review.

Small decision path

The rules engine is deterministic local code with no model call. Hosted latency still includes the network path between your agent and sipi.bot.