Datamimic – don't let your coding agent invent its own test world

Sep 16, 2026 11:58 AM - 1 hour ago 3

This repository contains the DATAMIMIC Community Edition (CE). MIT-licensed, Python-native, MCP-ready.

CE is afloat usable standalone for deterministic synthetic information procreation and PII-aware pseudonymization. The Enterprise Platform adds governed workflows, PII scanning, role-based access, audit logging, scheduling, multi-system execution, and the afloat operational furniture that regulated enterprises require.

👉 Enterprise Platform: datamimic.io | 📘 Docs: docs.datamimic.io | 📅 Book a strategy call: datamimic.io/contact

🤖 AI agent? Start astatine AGENTS.md and usage the task CLI: sphere caller intent arsenic model.dm.json, taxable an early champion effort via datamimic scaffold ... --format json, repair from the system issues, state an anticipation per stated requirement, and extremity connected verified=true. Existing earthy XML uses lint positive bounded dry-run.

Coverage Maintainability Python MIT MCP Ready

DATAMIMIC CE is the open-source deterministic information motor astatine the halfway of the DATAMIMIC Enterprise Platform. It is usable standalone for synthetic information procreation and PII-aware pseudonymization successful immoderate local, CI, aliases agent-driven workflow.

The Enterprise Platform adds the governed workflows, scanners, dashboards, and execution furniture that regulated enterprises require for production-scale test-data operations.

Available successful CE (this repo):

Generate afloat synthetic, deterministic datasets — model-driven, nary root information required
Pseudonymize staging/QA exports — deterministic (seeded) aliases privacy-maximized (non-seeded) section transformation; PII fields identified and modeled manually successful the XML pipeline
Execute single-system pipelines against PostgreSQL · MySQL · Oracle · MS SQL · SQLite · MongoDB · CSV · JSON · XML · XLSX · DbUnit · fixed-width (.fcw)
Model behavior — weighted authorities machines, composite multi-field references, power travel (<while>, <assert>), and a scriptable memstore for staged aggregation
Emit provenance — append-only execution logs and per-output contented hash for audit re-execution
Guide agents — machine-readable capabilities, progressive reference queries, and 1 canonical CLI scaffold transaction; an optional MCP adapter exposes the aforesaid authoring service

The Enterprise Platform adds:

PII scanner — probability-scored section discovery pinch configurable thresholds via DataWorkbench
Multi-system execution — Oracle / MongoDB / Kafka successful coordinated workflows pinch referential integrity
Industry connection templates — EDIFACT / SWIFT MT / HL7 v2.x / HL7 FHIR generated arsenic deterministic test/training artefacts
Governance layer — role-based dashboards, audit trails, support flows, reusable endeavor templates, scheduler
Performance core — Rust fastpath, ML/auto-regressive motor for analyzable distributions, keyset and manifest building, optimised distributed execution
On-premise / air-gapped deployment — podman-compose aliases Helm, pinch consulting-led rollout

Deployed successful regulated EU banking environments for deterministic trial information crossed Oracle, MongoDB, and Kafka pipelines. Reference customers disposable nether NDA — spot besides datamimic.io lawsuit studies.

AI agents: author, verify, and tally information models

The CLI is the baseline supplier contract. Install CE pinch pip instal datamimic-ce; inside this checkout, usage .venv/bin/datamimic truthful a old world installation cannot change the disposable schema aliases commands.

Need CLI tool Contract

Discover the unrecorded structural surface	datamimic capabilities	Compact machine-readable JSON scale by default; --full for the complete manifest, --section <name> for 1 section.
Learn the Intent Model progressively	datamimic reference authoring, past datamimic reference authoring --category <category> --kind <kind>	Start pinch the query catalogue, past load only the typed part needed.
Author a caller model	Preserve model.dm.json; tally datamimic scaffold model.dm.json --format json	One compile/lint/bounded-run/acceptance transaction per changed attempt. Stop connected verified=true; generated XML is runtime output.
Work pinch existing earthy XML	datamimic lint model.xml --format json, past datamimic dry-run model.xml --format json	Fix diagnostics, inspect bounded samples for intent, past usage datamimic tally model.xml only erstwhile existent execution is requested.
Find a DSL detail	datamimic reference overview, past a constrictive reference topic/name	Query the unrecorded exemplary and norm registries alternatively of guessing elements, generators, scope, distributions, aliases rules.

capabilities, authoring-reference projections, and the commands shown with --format json return machine-readable JSON. On a grounded scaffold attempt, change model.dm.json utilizing its system validation issues, typed repair, aliases rule diagnostics earlier retrying. A typed max_count remediation alternatively changes only the bounded scaffold parameter to astatine slightest its reported minimum. Never repetition an identical grounded call. A successful scaffold consequence is terminal for authoring, so do not lint aliases dry-run its generated XML again. Exact root fragments are discoverable done queries specified arsenic --category root --kind memstore.

When the calling situation already exposes DATAMIMIC MCP tools, they representation to the same canonical contracts and implementations: reference → datamimic_reference, scaffold → datamimic_scaffold, lint → datamimic_check, and dry-run → datamimic_run. Install the adapter pinch pip instal "datamimic-ce[mcp]"; registration specifications beryllium successful the MCP quickstart, not successful the authoring workflow. The adapter intentionally exposes only the four canonical reference, scaffold, check, and bounded-run operations; domain generation remains a Python/CLI capacity alternatively than a parallel MCP authoring path.