Pricing

What you pay for.

One simulation against a demographically grounded population of real American adults. Sandbox stays free. Studio is $99 a month. Enterprise is what you negotiate.

Sandbox

For building against the API.

per month — no card required

sk_test_* keys only

Included runs / month: 100
Agents per run: 100
Rate limit: 60 req/min
Result retention: 7 days
Webhook endpoints: 1
Data plane: Sandbox — synthetic + cached
Training Center: Read-only sample
Support: Community — GitHub issues
SLA: None

Building a prototype, reading the docs, running a few questions against a real-population simulation to see the shape of a response.

Get a sandbox key

Most chosen

Studio

For teams shipping production integrations.

$99

per month — $0.10 per run over 1,000

sk_live_* keys

Included runs / month: 1,000
Overage rate: $0.10 / run
Agents per run: 1,000
Rate limit: 1,000 req/min
Result retention: 90 days
Webhook endpoints: 10
Data plane: Production — Postgres + Redis + R2
Training Center: Full access
Support: Email — 24h business-day response
SLA: 99.9% — credits per terms

Product teams in production. The 1,000-run allowance covers normal weekly cadence for a small team; overage rolls onto the next invoice at a flat rate.

Start with Studio

Enterprise

For contracted buyers and research platforms.

Custom

annual contract — talk to founders for a quote

sk_live_* keys + dedicated infrastructure

Included runs / month: Contracted minimum
Agents per run: Negotiated — typical 10,000+
Rate limit: 5,000 req/min default
Result retention: Up to 1 year
Webhook endpoints: Unlimited
Data plane: Dedicated Redis + R2 + optional EU residency
Custom workflows: Private workflow submission
Training Center: Full + dedicated training runs
Support: Dedicated Slack — 4h response window
SLA: 99.95% — credits per contract
Onboarding: White-glove — founders on the call

Large research platforms, audience-intelligence buyers, regulated industries, and political or advocacy shops that need contracted SLAs and white-glove onboarding.

Talk to founders

How we charge

Per run, not per seat.

A run is one call to POST /v1/simulations. The base plan covers 1,000 runs a month — typical for a small product team. Beyond 1,000, runs roll into the next invoice at a flat $0.10 each. No per-seat fees, no overage tiers, no surprise bills. Below is the indicative cost split inside a single Studio run.

Cost component	What it covers	Share
Population grounding	Authoritative U.S. demographic and labor-force data sources. Refreshed quarterly. Source list available under NDA.	~30%
Workers AI inference	Cloudflare-hosted llama-3.3-70b for the influencer tier, llama-3.1-8b for worker agents. Per-token pass-through.	~55%
Platform — storage, edge, support	Neon Postgres + Upstash Redis + Cloudflare R2 + Railway compute + a thin margin to keep the lights on.	~15%

The split shifts when a run carries more agents — agents drive inference cost. A Studio run capped at 1,000 agents costs us a defensible fraction of $0.10; the margin pays for the parts of the system the bill does not (uptime, the on-call rotation, the auto-publish ledger, the benchmark page).

What a run answers

Sandbox and Studio runs are answered by one of our 76 calibrated workflows. Each workflow has a fixed KPI question backed by published research from authoritative sources we audit (full source list available under NDA). The auto-router maps your natural-language question to the closest workflow, and the population is asked the workflow's KPI question — not your literal phrasing. For literal-question mode, where agents are asked your exact wording, see Freeform Mode — arriving v1.1.

Frequently asked

Six questions, in the order we get them.

What happens if I hit the rate limit?: The API returns a 429 with a canonical Retry-After header. The official SDKs honor it with exponential backoff; if you are rolling your own client, the header is in seconds and we suggest a jittered retry. Sandbox keys are capped at 60 req/min, Studio at 1,000 req/min, Enterprise at 5,000 req/min by default.
Is there a hard cap on overage?: No hard cap, but a soft one. The first 30 days of overage attach a warning to the response envelope and send an email to the billing contact. From day 31 the agent-count ceiling becomes a 422 until the plan is upgraded or the invoice settles. Calling GET /v1/usage returns the live counter, and webhooks fire on threshold crossings so dashboards do not have to poll.
Is the Training Center included with Studio?: Yes — full access. The Training Center is the operator surface for inspecting calibration runs, sample agents, and the auto-publish feed. Sandbox keys get a read-only sample so prospective customers can see the shape of it; Studio and Enterprise customers get the full surface scoped to their organisation.
What does the Enterprise SLA actually cover?: 99.95% monthly availability calculated on the customer's contracted endpoints, measured at our edge. The contract specifies the credit schedule (typically 10% off the monthly invoice per 0.1 percentage point of miss, capped at 50%). The status page publishes incident reports — same correlation_id surface our on-call uses — so the math is auditable on both sides.
What payment methods do you accept?: Studio is Stripe — credit and debit cards plus US ACH for annual contracts; the Stripe Customer Portal handles upgrades, downgrades, and invoice history. Enterprise is wire or ACH against a PO; net-30 standard, net-60 on request with approved credit. We do not accept crypto.
Do you offer refunds?: Studio plans are pro-rated on cancellation within the first 30 days — no questions asked, email founders@simulix.com and the credit posts within a business day. Enterprise contracts follow the termination clause in the signed agreement. We do not pro-rate overage runs that have already been billed; those are paid for compute we have already run.

Something not covered here? Write to founders@simulix.com. Replies usually within a business day.