Tools Reference

Overview of all DataGrout core tools

DataGrout tools are automatically available on every server. They use this naming convention:

data-grout@{version}/{category}.{tool}@{tool_version}

For example: data-grout@1/discovery.discover@1

Tool Categories

Discovery Tools

Tools for finding and executing integration tools semantically.

Tool Purpose
discovery.discover@1 Find tools by natural language query or goal
discovery.plan@1 Generate a verified multi-step plan
discovery.guide@1 Interactive step-by-step workflow builder
discovery.perform@1 Execute a tool by name through the intelligence layer

See Discovery Tools for parameters and examples.

Data Tools

Pure JSON/structure manipulation. No LLM, no credits, no external calls.

Tool Purpose
data.get@1 Access a value at a path within a nested structure
data.pick@1 Keep only specified keys from a map or list of maps
data.omit@1 Remove specified keys from a map or list of maps
data.take@1 Return the first N items from an array
data.drop@1 Skip the first N items and return the rest
data.keys@1 Return keys of a map or indices of an array
data.count@1 Count items in an array, keys in a map, or characters in a string
data.flatten@1 Flatten a nested map into dot-path keys
data.merge@1 Merge two maps together
data.filter@1 Filter array items using declarative predicates
data.sort@1 Sort any array by value or by field specs
data.unique@1 Deduplicate an array by value or field

All Data tools accept either inline payload or a cache_ref from a prior tool response.

See Data Tools for parameters and examples.

Prism Tools

Data transformation, analysis, rendering, and code operations.

Tool Purpose
prism.refract@1 Transform data using natural language instructions
prism.focus@1 Convert between Semio semantic types
prism.chart@1 Generate charts from data
prism.render@1 Render data in various formats
prism.export@1 Export data to files
prism.paginate@1 Page through large datasets
prism.code_lens@1 Analyze code structure
prism.diff_analyzer@1 Compare code changes
prism.code_query@1 Query code semantically

See Prism Tools for parameters and examples.

Logic Tools

Symbolic memory for persistent agent knowledge.

Tool Purpose
logic.remember@1 Store facts in your Logic Cell
logic.query@1 Query stored knowledge
logic.constrain@1 Add rules and constraints
logic.forget@1 Remove stored facts
logic.reflect@1 Summarize what you know

See Logic Tools for parameters and examples.

Flow & Inspect Tools

Workflow orchestration, human-in-the-loop gates, and execution inspection.

Tool Purpose
flow.into@1 Define and execute a multi-step workflow
flow.route@1 Conditional dispatch with fall-through predicate matching
flow.request-approval@1 Pause workflow for human approval
flow.request-feedback@1 Collect missing data from a user mid-workflow
inspect.execution-history@1 List recent executions
inspect.execution-details@1 Get full details for a specific run
inspect.ctc-executions@1 View CTC-verified executions for a skill

See Flow and Inspect Tools for parameters and examples.

Warden Tools

Prompt injection detection and adversarial content analysis.

Tool Base Cost (low/med/high) Purpose
warden.canary@1 5 / 10 / 20 + LLM Protocol adherence check via dual-pass canary
warden.intent@1 5 / 10 / 20 + LLM Semantic intent lens — detects goal divergence and authority claims
warden.adjudicate@1 5 / 10 / 20 + LLM Adversarial fact/rule evaluation with Prolog engine
warden.ensemble@1 12 / 24 / 48 + LLM All three tiers in one call with weighted scoring

See Warden Tools for parameters and examples.

Math Tools

Deterministic numeric generation, statistical analysis, and modeling. No LLM, no credits.

Tool Purpose
math.range@1 Evenly-spaced sequence from start to stop by step
math.linspace@1 Exactly N points between start and stop
math.sequence@1 Named sequences: fibonacci, primes, geometric, etc.
math.sample@1 Random samples from uniform, normal, or exponential distributions
math.interpolate@1 Lerp, clamp, remap, smoothstep, and easing curves
math.describe@1 Descriptive statistics: mean, median, std, percentiles, histogram
math.window@1 Sliding window ops: moving average, cumulative sum, diff, pct_change, lag, EWMA
math.correlate@1 Pearson and Spearman pairwise correlation
math.trend@1 Regression fitting: linear, polynomial, exponential, logarithmic with forecast
math.normalize@1 Scale data: z-score, min-max, or percentile rank normalization
math.outliers@1 Detect outliers using IQR or z-score methods
math.rank@1 Rank values: ordinal, dense, average, or percentile

Analysis tools (describe, window, correlate, trend, normalize, outliers, rank) accept payload/cache_ref, composing naturally with Data and Frame tool outputs.

See Math Tools for parameters and examples.

Frame Tools

Columnar data operations on lists of records. No LLM, no credits.

Tool Purpose
frame.select@1 Keep, rename, or drop columns
frame.filter@1 Filter rows by declarative conditions
frame.sort@1 Sort by one or more fields
frame.group@1 Group and aggregate
frame.pivot@1 Reshape long data to wide format
frame.slice@1 Paginate or take top-N
frame.join@1 Join two record lists on shared keys

Frame tools accept cache_ref outputs from auto-paginated integration tools, enabling large dataset operations without passing data through the LLM context.

See Frame Tools for parameters and examples.

Scheduler Tools

Deferred and recurring tool execution.

Tool Purpose
scheduler.create@1 Schedule a tool call or workflow (time-based or event-driven)
scheduler.list@1 List active scheduled tasks
scheduler.cancel@1 Cancel a scheduled task

You can also pass a schedule argument directly to discovery.perform or flow.into without calling scheduler.create explicitly.

See Scheduler Tools for parameters and examples.

Naming Convention

All tools follow the namespace.tool@version pattern within the data-grout@1 scope:

  • discovery.* – Finding, planning, and executing tools
  • data.* – Pure JSON/structure manipulation
  • prism.* – Data transformation and code analysis
  • logic.* – Symbolic memory operations
  • flow.* – Workflow orchestration
  • inspect.* – Execution history and observability
  • warden.* – Adversarial content detection
  • math.* – Deterministic numeric operations
  • frame.* – Columnar data operations
  • scheduler.* – Deferred and recurring execution

Integration tools follow: {integration}@{version}/{tool}@{version} (e.g. salesforce@1/get_leads@1).

Many AI clients (Claude, ChatGPT, n8n, LangChain, Cursor) restrict function names to [a-zA-Z0-9_-]. DataGrout auto-detects these clients and serves flat names like salesforce_get_leads instead. You can call tools using either format. See Tool Naming and Compatibility for the full resolution chain and supported formats.

Error Format

All tools return errors in this structure:

{
  "error": {
    "code": "tool_not_found",
    "message": "Tool 'salesforce@1/nonexistent@1' not found"
  }
}

Common codes: tool_not_found, invalid_arguments, upstream_error, timeout, rate_limit, policy_violation.

Related