Pricing

One seat. $10 a month. Every feature.

No feature tiers to decode and no capability held back to force an upgrade. The market’s own alternatives are a $149 desktop add-in or a $10,000-a-seat analyst tool; this is the engine in between. AI employees are the one thing billed separately, and this page explains exactly how.

Pro

Most teams

$10 /user/mo

The single self-serve seat. Every standard feature, unlimited projects.

Join the waitlist

Enterprise

Custom

Deployment, compliance, and support commitments for pharma and federally funded programs.

Talk to us

Prices are in USD and exclusive of tax. Seats are billed monthly through Stripe.

What the $10 seat actually includes

This list is the entitlement grant the product enforces, not a marketing wish list. Everything below is built and working today.

The scheduling engine

  • Unlimited projects
  • Critical Path Method — the full dependency network, total float, and a ranked critical path
  • Monte Carlo simulation up to 50,000 iterations per run, reporting P50 / P80 / P90 with tornado sensitivity
  • Theory of Constraints with Drum-Buffer-Rope — critical chain, project and feeding buffers, fever chart
  • Risk register that drives the simulation, and retroactive rescheduling when a risk actually occurs
  • Interactive Gantt, plus schedule import from CSV or XLSX

Decisions and prioritization

  • Decision gates as first-class schedule objects — Go / No-Go / Pivot / Defer, with named owners and co-signs
  • WSJF and Cost of Delay
  • Portfolio dashboard across every program in the organization

Governance and access

  • Two-factor authentication (TOTP) with one-time recovery codes
  • Append-only, hash-chained audit log with in-app integrity verification
  • API access
  • The methodology coach and the schedule-aware copilot (the AI usage itself is metered separately — see below)

Not in Pro: SAML single sign-on, an on-premise deployment, SOC 2 Type II, a HIPAA BAA, and 21 CFR Part 11. Those are Enterprise, and most of them are roadmap — see the next section, where they are labeled as such.

Enterprise — custom

Sales-led, priced per program. It exists for the deployment, compliance, and support commitments a regulated or federally funded program needs in writing.

Included today

  • Everything in Pro
  • A dedicated customer success manager, onboarding, and training
  • Custom integrations for pharma and federally funded programs (DARPA, ARPA-E, ARPA-H, NIH)
  • Shipped today on the security side: 2FA/MFA, role-based access control, and the tamper-evident audit trail

On the roadmap — not shipped

Talk to us about timing before you commit a date to a regulator, a sponsor, or a board.

  • On-premise deployment option
  • SOC 2 Type II
  • SAML single sign-on (WorkOS)
  • HIPAA Business Associate Agreement
  • 21 CFR Part 11 electronic signatures — the tamper-evident audit trail is the foundation, and the foundation is all that ships today
Talk to us about Enterprise

How AI employees are billed

This is the part most AI products are vague about, so here it is in full. AI employees are billed separately from the seat, metered by usage: the estimated token cost of each completed AI job, plus a platform margin. Not credits. Not seats. Not hours.

Billed separately from the seat.

The $10 seat buys the platform. AI employees are not bundled into it and are not a per-tier entitlement — they are metered on their own line, so a team that never runs one never pays for one.

The unit is one completed AI job. Not one hour.

A job is one agent run, or one assistant request: the methodology coach, the Agent Builder, the skill-author wizard, AI work-breakdown decompose, TOC diagram generation, or the gate reviewer. A job may make several model calls internally; their tokens are summed and metered once, when the job completes. A run that fails is not metered at all. There is no time dimension in the meter — no hourly rate, no minimum, and nothing accrues while an AI employee sits idle.

The price of a job is its estimated token cost, plus a platform margin.

We total the tokens the job used, price them from a maintained provider price table, and add a stated margin. That is the entire formula. Being precise about one word: the base is our estimate of what those tokens cost, not the provider's own invoice line for them — see below.

No credits. Nothing to buy up front.

There is no credit pack, no prepaid balance, no expiring allowance, and no overage multiplier. This is a deliberate choice: unpredictable credit billing is the loudest complaint against the AI-agent platforms we track, and we would rather be boringly predictable than clever.

What we will and will not claim about the number.

The cost of a job is an estimate from a maintained provider price table, not the provider's own billed amount. The table is kept deliberately conservative, so when a provider is running promotional pricing our estimate can sit above what they actually charged us. We publish no worked dollar example, because we have no measured customer usage to base one on and will not invent one. Inside the product, your billing page shows raw model cost and billable usage side by side for the current month.

Nothing is metered unless metering is configured.

AI usage only bills when the usage meter is configured for your organization. Autonomous AI-employee runs are off during the invite-only beta and unlock at general availability.

Why we built it this way: every AI-agent platform we track bills in opaque credit packs, and unpredictable credit spend is their users’ single loudest complaint. Metering the real cost of the real call, and adding one stated margin, is the only model we could defend to a program leader who has to forecast it.

What counts as a seat

  • A seat is one active member of your organization. The minimum is one seat.
  • View-only guests admitted through a project share link are never billed. Showing a schedule to an outside stakeholder costs you nothing — billing for it would make sharing something you avoid, which is exactly backwards.
  • Deactivating or removing a member drops their seat. The seat count re-syncs with your membership and the change is prorated.
  • AI employees are resources, not members. They hold no seat, no login, and no access rights, so they never add to the seat count.

What you can actually do today

CritPath AI is in invite-only beta.

Public self-serve signup is closed and card checkout is not open to the public yet. You cannot buy a seat on this page today — and we would rather tell you that than show you a button that goes nowhere.

The waitlist is the way in.

Join it and we will get you into the beta. Invited organizations get the full scheduling engine and the AI assistance features; whoever signs up becomes the organization owner and can invite the rest of the team.

Already invited?

Sign in with the account from your invitation email. Nothing on this page changes for you.

Enterprise, pharma, or a federal program?

Talk to us directly. Deployment, compliance timing, and integration scope are a conversation, not a checkout.

Join the waitlistSign inBook a demo

Pricing questions

What exactly counts as a billable seat?

One active member of your organization. Deactivated and removed members stop counting. View-only guests admitted through a project share link never count — that carve-out is deliberate, so showing a schedule to an external stakeholder never costs you anything. AI employees are resources rather than members, so they hold no seat either. The minimum subscription is one seat, and the quantity re-syncs, prorated, when your membership changes.

Do I have to buy AI credits before I can use an AI employee?

No. There is no credit pack, no prepaid balance, and no expiring allowance to manage. AI usage is metered after the fact, per completed AI job, at the estimated token cost of the model calls it made plus a stated platform margin.

Does the $10 Pro seat include SAML single sign-on?

No. SAML SSO is an Enterprise capability, and it is on the roadmap rather than shipped — ask us about timing before you commit to a date. What Pro does include for account security today is two-factor authentication (TOTP) with one-time recovery codes, role-based access control, and the append-only audit log.

Do you charge extra for more projects, larger simulations, or more risks?

No. Pro is unlimited projects, and Monte Carlo runs of up to 50,000 iterations are included in the seat. Nothing in the scheduling engine is metered. AI usage is the only metered thing on the platform.

Can I buy a seat right now?

Not yet. CritPath AI is in invite-only beta: public self-serve registration and card checkout are both closed. Join the waitlist and we will get you in. When general availability opens, this page gets a real checkout path.

Is there a free plan?

No, and we would rather not dress one up. What other tools call a free tier is, here, simply the unsubscribed state of an organization. The honest answer today is the invite-only beta: join the waitlist, and beta organizations get the full scheduling engine and the AI assistance features, with autonomous agent runs held until general availability.

More questions about the product itself? Read the full FAQ, or see how the AI is architected and governed.

Ready to model your next milestone?

Join the waitlist and we’ll get you into the beta.

Join the waitlist