mloda 0.11.2 is out: free and open source under Apache-2.0 →
mloda
Free on GitHub

Connect your AI to all data it needs, in one entry point.

An open-source library that lets AI agents declare what data they need - and resolves how to fetch it, with data pipelines that are traceable, reproducible, and scalable.

Talk to usFree on GitHub

See it live:
mloda-backed
credit risk agent.

The agent asks for credit risk scores. mloda pools the data sources, runs the calculations, and returns scored results - with the full audit trail attached.

Step 1: The agent asks for the scoreStep 2: mloda gives back the scoresStep 3: How each score was made
Feature

Multi-modal data sources,
easily connected

Data lakes, warehouses, SQL stores - queried in place, no copies.

Dataframes, CSV, tables - native on Pandas, Polars, Spark.

Text, Markdown, PDF - parsed and ready for retrieval.

Third-party and internal services - wrapped once, reused everywhere.

mlodaVisit us on GitHub →

Your AI agent

Receives exactly the data it asked for - resolved, processed, and traced to source.

CSVJSONParquetFeatherORCPlain textMarkdownYAMLSQLiteAPI / runtime data
Feature

Plug and play data plugins -
define, chain, swap

Compose data pipelines like words → Live demo

income
__sum_aggr
age
__avg_aggr
user_query
__injection_checked
__retrieved
__pii_redacted

Adjust and reuse what's already there, no rewriting → Live demo

✕
linear pipeline - rewritten every time
✓
mloda - swap one plugin, rest untouched

Use cases

Let agents use knowledge graphs
open-kgo · Knowledge graphs

Let agents use knowledge graphs

An agent answers questions across knowledge sources - company graphs, software inventories, even its own memory - with one simple request. Runs fully offline: nothing leaves your machine.

Access case →
A document chatbot, declared in one line
rag_integration · RAG pipelines

A document chatbot, declared in one line

Point it at your documents: personal data is removed, texts are split and prepared, and the right passages reach your chatbot. Every step is a plugin you can swap.

Access case →
Public-sector data, agent-ready
govdata · Open government data

Public-sector data, agent-ready

German open government data made usable for AI agents - ask for a dataset, get it clean and traceable. Funded by the German Federal Ministry of Research, Technology and Space via the Prototype Fund.

Access case →
A credit risk agent, fully auditable
credit risk · Financial services

A credit risk agent, fully auditable

An AI agent asks for credit risk scores - mloda pools the data sources, runs the calculations, and returns scores with a full audit trail for every value.

Access case →
Credit risk agentClimate risk agentCustomer service agentRAG systemContext assemblyOpenclaw-like agent

Get started

Three ways in - install the library, explore the source, or let your coding agent wire it up. Same library, from notebook to production.

Install it yourself

Copy one command - pip install mloda and declare your first Feature in minutes.

pip install mloda
Click to copy the install command.

Explore the source

Star the repo, read the code, open an issue - mloda is built in the open.

Let your coding agent do it

Copy one prompt - your coding agent installs mloda and wires it into your project.

Wire in mloda!
Click to copy a prompt for your coding agent.

mloda fits into your existing stack - connect, ask, get traced data.
Three steps to wire it in.

PostgresS3 / filesREST APIFeatureGroupplugin, not rewrite
1

Connect what you already have

Wrap your DB, API, or files in a FeatureGroup - a small plugin, not a rewrite. Set it up once, reuse it everywhere.

your agentrequest = ["customer_ltv","churn_risk__scored"]__suffix = optional transform
2

Let your agent ask

Agents declare features by name - no pipeline code, no glue. Need processing? Add a suffix: __scored, __aggregated, and more.

mlodatracedPandas·Spark·Polars
3

Get data, fully traced

mloda resolves the dependency chain on your compute engine and returns results with lineage built in.

mloda

Start building with us today.

Talk to usExplore the open source

Open source

Free to use. The library, all steps that ship with it, and the sets built on it. Apache-2.0.

Plugin Bundles

Curated connector and transform packs for common stacks.

Coming soon

Custom

Pay to control. We build and support your steps and connections, and set up who may run which step, and for every number what ran, which version, and what it read. Contact us.