Skip to main content

AI Pricing Transparency FAQ

Use this FAQ in Stainless-transition, OpenAPI-to-MCP, and GrayMatter pilot conversations when the buyer wants to understand what drives price before they commit. The short version: subscription pays for access and baseline platform capability; credits pay for variable AI, memory, workflow, and hosted-runtime work that should be visible instead of hidden.

What Am I Paying For?​

ValkyrAI pricing has three practical layers:

  • Platform access: account, workspace, generated app, workflow, and admin surfaces included in the selected plan.
  • Credits: metered work such as model calls, GrayMatter retrieval/write operations, generation jobs, MCP-published service calls, and other billable automations.
  • Hosted runtime or support add-ons: dedicated infrastructure, isolated instances, implementation help, or custom support when the buyer wants Valkyr Labs to operate more of the stack.

Plan pages and order forms should treat the live product catalog and credit pricing matrix as the source of truth. Sales notes, demo decks, and older launch copy are only directional unless they point back to the current catalog.

Why Use Credits Instead of Hiding AI Costs in a Flat Fee?​

AI usage has real variance. One customer may use GrayMatter for light recall and a few workflow runs; another may generate apps, run large memory searches, attach media, and publish MCP services every day. A flat fee that hides those costs either overcharges quiet teams or quietly breaks margin for active teams.

Credits make that variance explicit. They also let teams approve spend before high-volume work, attribute cost to projects, and compare the value of an automation against its usage.

What Should Be Visible Before Checkout?​

The checkout or pilot proposal should make these items clear:

  • Which plan or pilot package unlocks the requested capability.
  • Whether credits are included, prepaid, or purchased separately.
  • Which actions consume credits.
  • Whether hosted runtime costs are included or billed as a separate add-on.
  • What happens when credits run low, including pause, upgrade, or top-up behavior.
  • Which limits are technical guardrails versus commercial package limits.

If any of those answers are unknown, the right answer is "pending catalog confirmation," not an invented discount or estimated price presented as final.

How Do GrayMatter Costs Work?​

GrayMatter cost is driven by memory operations and context assembly:

  • writing durable memory entries
  • semantic indexing and reindexing
  • retrieval and context hydration
  • long-context assembly for agents and workflows
  • compaction, pruning, and other memory maintenance jobs

Operators should use the GrayMatter memory dashboard and credit ledger to inspect burn rate, hit quality, and recommendations before scaling a workflow. Better memory hygiene should lower waste; it should not become a hidden tax.

How Do ValkyrAI Workflow Costs Work?​

Workflow cost depends on the modules involved:

  • simple generated-object reads and writes may be plan-covered or low-cost
  • LLM calls consume model and context resources
  • media, file, deployment, and generation steps may have additional runtime costs
  • MCP-published service calls can have per-call pricing when the service owner enables monetization

For production workflows, estimate the expected run frequency and the expensive steps first. Then use credits and usage receipts to validate the estimate during the pilot.

How Should We Explain Hosting?​

Hosted runtime pricing should be discussed separately from AI usage:

  • Shared control-plane access covers ValkyrAI platform surfaces and generated object workflows.
  • Managed or isolated runtime adds infrastructure and operator responsibility.
  • Dedicated, app-database, or BYOC deployments may require explicit support and runtime add-ons.

The buyer should know whether they are buying software access, managed infrastructure, implementation support, or all three. Those are different commitments.

What About Stainless-Transition Buyers?​

Do not position Valkyr as a drop-in Stainless replacement. The useful pricing conversation is narrower and stronger:

  • If the buyer needs API-first app generation, ThorAPI/ValkyrAI can reduce custom build work.
  • If the buyer needs agent memory, GrayMatter adds continuity and provenance beyond SDK generation.
  • If the buyer needs MCP action surfaces, ValkyrAI can expose controlled workflow tools with RBAC and audit boundaries.
  • If the buyer only needs polished multi-language SDK output, scope that honestly before quoting.

Price the pilot around the actual problem: spec-to-stack generation, memory continuity, workflow automation, or hosted runtime operation.

What Should Sales Avoid Saying?​

Avoid these claims unless the live catalog, current implementation, and order form support them:

  • "Unlimited AI usage"
  • "All hosting included"
  • "No per-use costs"
  • "Drop-in Stainless replacement"
  • "Every SDK language is first-class"
  • "Credits are only a billing detail"

Clean pricing builds trust. Surprise metering destroys it quickly, and not in a fun dramatic way.

What Evidence Should We Bring To A Pricing Call?​

Bring links or screenshots for:

  • the current plan or product catalog entry
  • the credit pricing matrix or buyer-facing credit explanation
  • usage receipts from a similar workflow or pilot
  • the GrayMatter memory dashboard when memory cost is part of the deal
  • the deployment/runtime option being proposed
  • the scoped acceptance criteria for the pilot

The goal is not to make pricing look artificially simple. The goal is to make it inspectable enough that a serious buyer can say yes without feeling trapped.