Tool Pricing

Transparent, per-call pricing. Every call pays the gateway base; deterministic tools add nothing on top. Intelligence tools charge a fixed platform premium plus actual LLM cost.

All values in DataGrout credits · LLM costs added at runtime

Gateway Base

1
Infrastructure cost on every tool call — auth, materialization, policy, cadence and receipts. Never waived.

Tool Premium

0–48
Fixed orchestration value for the tool itself. Depends on complexity and effort level.

LLM Passthrough

actual × 1.2
Actual token cost with 1.2× billing margin. Zero for BYOK users and deterministic tools.

discovery.perform billing

When tools are called through discovery.perform, the inner tool's own price is used — not a flat wrapper fee. Deterministic inner tools still pay the gateway base; perform is not a discount. Non-deterministic inner tools are billed at the Credits value shown in the tables below, plus any LLM passthrough from actual token usage.

inner_tool_credits + LLM passthrough

Deterministic

No premium

Pure data shaping and inspection. No premium above the gateway base, and no LLM cost.

Tool Description Credits + LLM
data.* JSON transforms, aggregation, mapping, and filtering 1 —
frame.* Record-oriented grouping, joins, pivots, and slicing 1 —
signal.* Deterministic signal processing NIFs — FFT, filters, resampling 1 —
linalg.* Deterministic linear algebra NIFs — matmul, decompositions, solvers 1 —
math.* Deterministic statistics, modeling, and numeric utilities 1 —
ephemerals.list / inspect Browse active cached result sets 1 —
deliverables.list / get Retrieve persistent work products 1 —
toolsmith.catalog Browse saved skills and invocation hints 1 —

Discovery

Find, plan, and execute the right tool for a goal.

Tool Description Credits + LLM
discovery.summary Capability overview across connected tools 3 varies
discovery.discover Semantic search with ranked matches 3 varies
discovery.plan Verified multi-step plan from a goal 3 varies
discovery.guide Interactive guided exploration 3 varies
discovery.perform Execute a selected tool or saved skill inner inner

Prism

AI-assisted transforms, analysis, rendering, and visualization.

Tool Description Credits + LLM
prism.refract Natural-language data transformation 5 1–7
prism.analyze Structured reasoning over payloads 3 / 5 +LLM
prism.chart Chart generation and cached chart reuse 5 / 1 1–4 / —
prism.focus / render / export / paginate Format conversion, focused extraction, and pagination helpers varies varies

Latent

Conceptual expansion, angular navigation, and boundary discovery.

Tool Description Credits + LLM
latent.expand Expand a seed concept via depth, breadth, or bridge 2–5 1–5
latent.orient / horizon Angular navigation and boundary exploration 8 2–14

Invariant

Semantic code analysis and alignment checking.

Tool Description Credits + LLM
invariant.code_lens Extract structural and semantic code facts 2 / 4 — / 0–1
invariant.code_query Query stored code facts deterministically 3 —
invariant.diff_analyzer Check changes against an intended goal 4 0–1

Logic

Persistent symbolic memory and reasoning.

Tool Description Credits + LLM
logic.remember / assert Store facts in natural language or structured form 1–3 optional
logic.query / hydrate / reflect Retrieve, assemble, and summarize relevant knowledge 1–3 optional
logic.constrain / forget Manage rules and remove outdated facts 1–3 optional
logic.worlds / export / import / tabulate Namespace, portability, and tabular views 1 —

Warden

Safety checks and adversarial-content review.

Tool Description Credits + LLM
warden.canary Protocol and instruction integrity checks 5 / 10 / 20 +LLM
warden.intent Intent and authority analysis 5 / 10 / 20 +LLM
warden.adjudicate Threat classification with rule-grounded evidence 5 / 10 / 20 +LLM
warden.ensemble Combined verdict with weighted scoring 12 / 24 / 48 +LLM

Flow

Workflow orchestration, branching, and human gates.

Tool Description Credits + LLM
flow.into Execute a multi-step skill or inline plan 5 varies
flow.route Conditional branching and routing 5 varies
flow.request-approval Human approval checkpoint 2 —
flow.request-feedback Collect missing input mid-run 2 —

Toolsmith

Create, refine, catalog, and invoke reusable skills.

Tool Description Credits + LLM
toolsmith.forge Create a new saved skill from a goal 5 1–3
toolsmith.temper Refine or fork an existing skill 3 / 5 1–3
toolsmith.invoke Run a saved catalog skill by skill_id standard depends on skill

Tasks, Delivery, and Review

Durable handles, persistent outputs, and execution review.

Tool Description Credits + LLM
tasks.* Poll, wait on, fetch, list, or cancel background work 1 —
deliverables.register Persist a result as a reusable work product 2 —
inspect.* Review execution details, history, and certified runs 1 —

Scheduler and Latent

Recurring execution and conceptual expansion.

Tool Description Credits + LLM
scheduler.* Create, list, and cancel scheduled work standard optional

Docs

Persistent document storage for agents — design docs, memos, checklists.

Tool Description Credits + LLM
docs.create Create a new persistent document 1 —
docs.update Modify content, title, tags, or append 1 —
docs.get Retrieve a document by ref 1 —
docs.list List documents with optional tag filter 1 —

Library

No premium

Semantic search and retrieval over DataGrout documentation and papers. No premium above the gateway base.

Tool Description Credits + LLM
library.search Search docs and papers by natural language query 1 —
library.get Retrieve full documents or specific sections by ref 1 —

Attention and Batteries

Symbolic recall across sessions, and installable reasoning packs.

Tool Description Credits + LLM
attention.nearest / remember / curate Retrieve and record long-horizon memory across sessions 1 —
batteries.search / install Find and install prebuilt logic batteries into a cell 1 —

Agents and Workspaces

Agent orchestration, shared workspaces, and run history.

Tool Description Credits + LLM
agents.* Create and orchestrate agents over a goal 1 —
workspaces.* Shared context and artefacts for a group of agents 1 —
runs.* Inspect past orchestration runs and their steps 1 —

Grounded Research

Web search grounded in scraped sources, with citations.

Tool Description Credits + LLM
inference.search One grounded question — search, scrape, synthesise with citations 1 2–3
inference.research Batch research over many questions, compiled into a report 1 per question

Lumen

Token and cost telemetry captured at the traffic layer.

Tool Description Credits + LLM
lumen.dashboard Live usage, spend, and rolling metrics 1 —
lumen.laps Lap-based session boundaries and per-lap totals 1 —
lumen.compare Compare two laps or windows for cost and token deltas 1 —

Reactor

Expose a saved rule or skill as a callable HTTP endpoint.

Tool Description Credits + LLM
reactor.expose Publish a rule pack as an invokable endpoint 1 —
reactor.list / update / revoke Manage published endpoints and their access 1 —

Store

Durable key-value storage scoped to your account.

Tool Description Credits + LLM
store.put / get / delete Write, read, and remove stored values 1 —
store.query List and filter stored keys 1 —

Pipeline Add-ons

via perform

Bundled with discovery.perform — cheaper than standalone transforms.

Tool Description Credits + LLM
refract (pipeline) Natural-language transform as a perform parameter 3 +LLM
chart (pipeline) Chart generation as a perform parameter 3 +LLM
analyze (pipeline) Shallow or deep analysis as a perform parameter 1 / 3 +LLM

Smart Panels

Publish and consume reusable multi-tool workflows.

Tool Description Credits + LLM
Smart Panel publish One-time cost to publish a panel 25 —
Smart Panel use Per-use base cost for published panels 2 varies

Notes

  • Credits = platform premium only. LLM passthrough is added at runtime based on actual token usage.
  • BYOK (Bring Your Own Key) users pay premiums only — LLM costs go directly to the provider.
  • Reuse pricing — cached skills, charts, and code lens results cost ~1 credit vs. full generation price.
  • Pipeline add-ons — refract, chart, and analyze via discovery.perform are 2 credits cheaper than standalone calls.
  • Warden effort levels — low / medium / high determine how many model passes are used for safety evaluation.
  • Annotations & embeddings — comped by the platform (charged then credited, shown on receipts).
  • Connected integration tools (e.g. Salesforce, QuickBooks) — 1 credit base per call, no tool premium.
  • Encryption at rest — all Prolog facts (Logic, Invariant, Governor world model, workspace knowledge bases) are encrypted at rest using AES-256-GCM.