Transform Data.
Analyze. Visualize. Export.
Prism is DataGrout's data transformation, analysis, and visualization layer. Describe what you want in natural language — Prism generates verified, sandboxed code that runs deterministically on every subsequent call.
9 tools
refract · analyze · chart · focus · render · export · paginate · orient · horizon
Self-healing & cached
first call generates code · every repeat skips AI entirely
Inline on any tool
pipe refract or chart onto any discovery.perform result
Free to start · No credit card required
Nine Tools. Every Data Operation Your Agent Needs.
From natural language transforms to type bridging, charting, and paginated exports — Prism handles the full data lifecycle between your tools.
Natural Language Transform
Reshape any data structure using plain English. Describe the transformation — group, flatten, rename, filter, sum, merge — and Prism generates sandboxed executable code. The first call uses AI; every repeat with the same intent hits the cache and runs deterministically at near-zero cost.
Key capabilities
- Any JSON shape: arrays, nested maps, mixed structures
- Self-healing: failed edge cases auto-correct and re-cache
- Verbose mode returns transformation metadata
- Inline: add refract to any discovery.perform call
- Progressive efficiency: same intent = skip AI entirely
What Prism Solves
Every agent workflow generates data that needs reshaping, bridging, analyzing, or presenting. Prism handles all of it without leaving the platform.
Reshape Data Between Pipeline Steps
Agent retrieves raw Salesforce leads — 47 fields each. Prism.refract extracts just name, email, and annual revenue in one call. Cached on first run; every repeat skips AI and executes deterministically.
Related Solutions
Cross-System Type Bridging
Qualified Salesforce leads need to become QuickBooks customers. Prism.focus bridges crm.lead@1 → billing.customer@1 automatically — no manual field mapping, no glue code. discovery.plan inserts it for you.
Related Solutions
Instant Charts From Any Tool Output
Add a chart parameter to any discovery.perform call and get back a hosted SVG URL, sparkline, and summary statistics — no separate visualization step, no extra round-trip.
Related Solutions
Decision Support & Anomaly Detection
Feed a dataset to prism.analyze with mode: 'exploratory' and get back structured findings, evidence citations, and confidence scores — not just data, but reasoned conclusions.
Related Solutions
Automated Report Generation
Chain tool results into prism.render for branded reports in any format — markdown summaries, PDF decks, XLSX tables, or HTML dashboards — from a single natural language instruction.
Related Solutions
Progressive Efficiency Over Time
The first time an agent runs a Prism transform, AI generates verified executable code. Every subsequent call with the same intent runs from cache — zero AI cost, deterministic output. Agents get faster the more they run.
Related Solutions
Prism Amplifies Every Other DataGrout Tool
The real power of Prism is how it slots into the rest of the platform — transforming, analyzing, and presenting results from any tool without leaving your workflow.
Inline Transform on Any Tool Call
Add refract or chart as a parameter directly to any discovery.perform call. The result is automatically piped through Prism server-side before reaching your agent — no extra round-trip, no re-fetching the raw data.
✦ One call from raw tool output to formatted, visualized result
Automated Reporting Pipelines
Chain prism.render or prism.export as the final step of a flow.into plan. Every time the workflow runs — on a schedule or event trigger — it outputs a fresh PDF report or XLSX export automatically.
✦ Hands-free, repeatable report generation at any cadence
Filter First, Transform Cheaper
Use free Data or Frame tools to filter and slim a dataset before passing it to prism.refract or prism.analyze. Smaller payloads = lower AI costs and faster cache hits.
✦ 10–100x cheaper AI transforms by pre-filtering with free tools
Continuous Analytical Monitoring
Schedule a prism.analyze call via Governor to run on a recurring basis against live data. Governor's symbolic reflex layer checks preconditions for free; analysis only fires when the data actually changes.
✦ Always-on analysis that only burns tokens when it matters
Integrate Prism in Minutes
One import. Auth, retries, and caching all handled by the Conduit SDK.
from datagrout.conduit import Client
async with Client(
"https://gateway.datagrout.ai/servers/{uuid}/mcp",
auth={"bearer": "your-access-token"}
) as client:
# Transform raw invoice data
result = await client.perform("data-grout@1/prism.refract@1", {
"goal": "group invoices by customer and sum totals",
"payload": [
{"customer": "Acme", "amount": 500, "status": "paid"},
{"customer": "Globex", "amount": 300, "status": "unpaid"},
{"customer": "Acme", "amount": 200, "status": "unpaid"},
]
})
# result["result"] = [
# {"customer": "Acme", "total": 700, "count": 2},
# {"customer": "Globex", "total": 300, "count": 1}
# ]
# Use cache_ref from a prior tool call — data stays server-side
transformed = await client.perform("data-grout@1/prism.refract@1", {
"goal": "extract only name and email fields, flatten address to one line",
"payload": prior_cache_ref
})Also available in TypeScript, Rust, Elixir, and Ruby. View Conduit SDK →
Frequently Asked Questions
Common questions from developers integrating Prism into their agents and workflows.
More questions? Read the full Prism documentation.
Stop writing data glue. Start describing outcomes.
Prism transforms, bridges, analyzes, and visualizes your data — in natural language, with verified code, from any tool in your mesh.
Free to start · No credit card required