Pricing

HumanChain bills in credits. One credit = one human response from the marketplace, or one LLM fallback response. Nominated chains have flat-fee seat pricing; chain-internal queries don’t consume credits.

This page covers the v0.15 pricing model. Numbers below may move as we collect data on real-world usage patterns - the dashboard always shows the current rates.

The free tier

25 consults free every month. No credit card required - and it’s recurring: the allowance resets on the 1st of each month (UTC), not a one-time trial window.

What the free plan includes, and where it’s capped:

• 25 consult() calls per calendar month. Hitting the cap returns a HumanChainQuotaExceeded error with an upgrade prompt; the counter resets at the start of the next month.
• wait_seconds clamped to 120s (paid plans allow the full 1800s).
• Question framing is off. The paid route → frame → dig engine - which reframes your question so a human can answer it concretely - is a paid feature.
• Sandbox mode is unlimited and never counts against the 25.

The free tier is meant to be enough to wire HumanChain into your agent, run integration tests, and make a handful of real calls each month before deciding a paid plan is worth it.

Paid tier - pay per credit

Once your free tier is exhausted, you’re on pay-as-you-go. No monthly minimum, no contract, no commitment. Top up your account in $10 / $50 / $250 chunks; credits don’t expire.

Approximate per-credit pricing (subject to change as the marketplace calibrates):

Credit type	Cost (USD)
LLM fallback (Groq Llama 3.3 70B)	$0.02
Marketplace - general domains	$0.50
Marketplace - software / data	$0.75
Marketplace - legal / financial	$2.00
Marketplace - medical / regulatory	$3.50
Nominated-chain internal	$0.00 (covered by seat fee)

The Hub’s split: roughly 70% goes to the responding expert, 30% to HumanChain. Experts choose the domains they answer in; market rates emerge from supply and demand.

A typical pattern:

# 1 credit, mostly LLM fallback territory
consult(question="Is this Postgres migration safe?", ...)
# ≈ $0.50 if a human picks it up, $0.02 if LLM fallback

# 3 credits at legal market rate
consult(question="Approve this contract clause?", ...,
        target_responses=3, approval_method="consensus")
# ≈ $6.00 if 3 human lawyers answer, $0.06 if all 3 fall back

# 3 credits, medical rate, strict
consult(question="Approve this triage decision?",
        target_responses=3, approval_method="all_must_agree",
        require_human=True)
# ≈ $10.50 - high-stakes, all-human, expensive on purpose

The math is meant to be intuitive: cheaper for low-stakes, more expensive for high-stakes, with the expert getting most of the margin.

Nominated chains - seat pricing

Nominated chains have a different billing shape. You pay a flat fee per seat (master + nominated experts); chain-internal queries are free. Public-marketplace overflow is metered at marketplace rates if you enable it.

The current shape (v0.15 - pricing will likely shift as we onboard more chains):

Tier	What’s included	Approx. monthly
Starter chain	1 master + up to 5 experts, unlimited internal queries, sandbox mode	$99
Team chain	1 master + up to 25 experts, unlimited internal, expertise grading insights dashboard, basic SLA tracking	$499
Org chain	Unlimited seats, full expertise grading API, custom roles, audit log, SSO, on-call rotation hooks	Talk to us

Public-marketplace overflow on top of any tier is charged at the per-credit rates above (only if you enable allow_public_fallback=true).

Tip

Most teams underestimate how much value they get out of a starter chain. 5 experts answering each other’s agents’ questions is already a meaningful internal helpline. Try the $99 tier for a month before assuming you need the larger ones.

Sandbox mode - always free

HUMANCHAIN_MODE=sandbox returns synthetic responses without hitting the real backend. Use it for:

• CI test suites
• Local agent development
• Demos that don’t need real human answers
• Anything you don’t want billing for

Sandbox responses are structurally valid ConsultResponse objects. Your agent code paths exercise the same branches; nothing burns credits; no humans get woken. Free forever, no rate limits.

What doesn’t cost money

The list of things you can do on HumanChain without paying:

• Signing up (no card)
• 25 consults every month (no card, recurring)
• Sandbox mode (forever, no card)
• Reading the docs
• Inviting people to your chain
• Calibration battery for nominated experts
• Submitting feedback on expert answers (flag a bad answer; help expertise grading learn - actually saves you money over time)

What does cost money

The list of things that do consume credits / dollars:

• Real consult() calls past the monthly 25 (upgrade to a paid plan)
• Public-marketplace queries (always, after free tier)
• LLM fallback (cheap, but not free)
• Nominated-chain seat fees (after starter trial)

Billing transparency

Every consult() response carries the cost_credits field. You always know what the call cost, before you decide what to do with the response.

The dashboard shows:

• Credit balance and recent burn rate
• Cost breakdown by domain / by agent / by approval_method
• Top-spending consult IDs, in case one runaway loop is eating your budget
• Alerts when balance drops below a threshold you set

Add a card via the dashboard. Stripe-backed, charges only fire when you top up (or when auto-refill triggers, if you enable it).

What’s coming next

In v0.16:

• Stripe self-serve top-up with auto-refill thresholds
• Volume discounts for >10k credits / month
• Pre-paid annual commitments with a discount for nominated-chain customers
• Per-domain rate locking for enterprise: pin a rate for your team for the year regardless of market movement

The principle stays the same: pay-as-you-go, no contracts, no lock-in. Roll-overs and discounts are conveniences, not commitments.

Next step

The other operations page:

Security & privacy →