Kedro

Open-source Python framework for building maintainable and modular data science pipelines. Kedro applies software engineering best practices (modularity, versioning, configuration management, testing) to data and ML pipelines. Uses a node/pipeline/catalog abstraction: nodes are Python functions, pipelines are DAGs of nodes, and the catalog manages data I/O. Popular in enterprise data science for reproducibility and project structure standards.

Evaluated Mar 06, 2026 (0d ago) v0.19+
Homepage ↗ Repo ↗ Developer Tools data-pipeline python open-source dag mlops data-engineering reproducibility
⚙ Agent Friendliness
67
/ 100
Can an agent use this?
🔒 Security
82
/ 100
Is it safe for agents?
⚡ Reliability
82
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
--
Documentation
85
Error Messages
80
Auth Simplicity
100
Rate Limits
100

🔒 Security

TLS Enforcement
85
Auth Strength
80
Scope Granularity
75
Dep. Hygiene
88
Secret Handling
85

Local framework — no network exposure by default. Credentials managed via credentials.yaml with env var interpolation (keeps secrets out of code). Apache 2.0 license with active community security reviews. No external API calls from the framework itself.

⚡ Reliability

Uptime/SLA
90
Version Stability
82
Breaking Changes
78
Error Recovery
80
AF Security Reliability

Best When

You're building reusable, testable, versioned data science pipelines in Python and want project structure that scales from prototype to production.

Avoid When

You need managed scheduling, event triggers, or cloud UI — Kedro is a framework, not an orchestrator. Pair it with Airflow or Prefect for scheduling.

Use Cases

  • Structure agent data preprocessing pipelines with proper modular design, configuration management, and reproducibility baked in
  • Build ML feature engineering pipelines that can be run locally or deployed to Airflow/Prefect/Kubeflow with Kedro plugins
  • Create versioned datasets and model artifacts with Kedro's DataCatalog for tracking what data was used in each pipeline run
  • Generate pipeline visualizations and documentation automatically with Kedro-Viz for agent debugging and monitoring
  • Standardize multi-team data science projects with Kedro's project template and conventions to reduce agent hallucination risks

Not For

  • Event-driven or real-time pipelines — Kedro is batch-oriented; use Kestra or Temporal for event-driven workflows
  • Teams wanting a managed cloud service — Kedro is a local framework; use Prefect Cloud or Dagster Cloud for managed orchestration
  • Simple one-off scripts — Kedro's structure overhead is only justified for multi-node, reusable pipelines

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
No
SDK
Yes
Webhooks
No

Authentication

Methods: none
OAuth: No Scopes: No

Kedro is a local Python framework with no API auth. Credentials for data sources (S3, databases) are managed via Kedro's credentials.yaml with environment variable interpolation. No API key or auth for the framework itself.

Pricing

Model: open_source
Free tier: Yes
Requires CC: No

Kedro is completely free and open source. The ecosystem (plugins for Airflow, MLflow, etc.) is also open source. No vendor lock-in.

Agent Metadata

Pagination
none
Idempotent
Full
Retry Guidance
Not documented

Known Gotchas

  • Kedro has no REST API — agents must invoke pipelines via Python SDK or kedro run CLI command, not HTTP calls
  • credentials.yaml requires manual setup before first run — no auto-discovery of cloud credentials; agents must configure catalog and credentials explicitly
  • Kedro's DataCatalog uses YAML-defined datasets — agents generating pipelines must understand catalog config format to register new data sources
  • Pipeline visualization requires Kedro-Viz (separate install) — not included by default
  • Node function signatures must match catalog dataset names exactly — mismatches cause runtime errors that can be hard to debug
  • Kedro runner options (SequentialRunner, ParallelRunner, ThreadRunner) have different behavior for shared state — agents should use SequentialRunner for predictable execution
  • Kedro projects use a specific directory structure (conf/, data/, src/) — agents generating Kedro code must follow this structure or pipelines won't load

Alternatives

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Kedro.

$99

Scores are editorial opinions as of 2026-03-06.

5215
Packages Evaluated
26151
Need Evaluation
173
Need Re-evaluation
Community Powered