RoboSystems is live — what we built, where it's going, and how to get involved

[jfrench9]

Hi — I'm Joey, the founder of RoboSystems. This is our first Discussion post, so I want to use it to lay out what this project is, what's here today, and where we're headed.

What is RoboSystems?

RoboSystems is a financial knowledge graph platform. It takes structured financial data — starting with SEC XBRL filings — and transforms it into a queryable graph database with semantic enrichment, AI-native tooling, and a full deployment pipeline.

The core insight: XBRL is a graph, not a table. Entities file reports. Reports contain facts. Facts are qualified by elements, periods, units, and dimensions. Trying to flatten this into relational tables creates the complexity that's kept XBRL data practically unusable for 15 years. We model it as a graph and layer semantic enrichment on top — canonical concepts, disclosure classification via the Seattle Method, and confidence scoring from graph analytics.

The platform ships with:

• LadybugDB — An embedded columnar graph database purpose-built for financial analytics. No external server. Native DuckDB integration for staging and materialization.
• SEC XBRL Pipeline — A six-stage Dagster-orchestrated pipeline: Download → Process → Stage → Enrich → Materialize → Publish. 14 node types, 21 relationship types. 10,000+ reporting companies dating back to 2015.
• MCP Server — npx -y @robosystems/mcp gives Claude, Cursor, or any MCP-compatible client structured access to financial data. Tools like get-financial-statement, build-fact-grid, and resolve-element let AI query SEC filings without knowing XBRL or Cypher.
• Full AWS Infrastructure — CloudFormation templates for the complete stack: ECS Fargate API, EC2 auto-scaling LadybugDB clusters, RDS, ElastiCache, S3, DynamoDB registries. GitHub Actions CI/CD with OIDC auth.
• Client SDKs — TypeScript, Python, and MCP.

Everything is Apache 2.0.

What's here today

The repo is at v1.4.25 with 444 merged PRs and 74% test coverage. The SEC pipeline is production-grade — we run it on the managed platform at robosystems.ai. The local development experience is designed to get you from clone to queryable graph in minutes:

brew install uv just
git clone https://github.com/RoboFinSystems/robosystems.git
cd robosystems
just start
just demo-sec

That loads NVIDIA's SEC filings into a local LadybugDB instance, runs the enrichment pipeline, and gives you a working knowledge graph. From there you can query with Cypher (just graph-query sec "MATCH ..."), connect the MCP server to Claude Desktop, or use the Python/TypeScript SDKs.

The Wiki has detailed guides:

• Architecture Overview — Full system design
• SEC XBRL Pipeline — Loading and querying SEC data
• Bootstrap Guide — Deploying to your own AWS account
• Accounting Demo — Graph-based double-entry bookkeeping

The ecosystem

RoboSystems is the backend. The broader ecosystem includes:

Repository	Purpose
robosystems	Core API, graph database, pipelines, infrastructure
robosystems-app	Main platform dashboard (Next.js)
roboledger-app	Accounting and financial management (Next.js)
roboinvestor-app	Investment analysis and portfolio tools (Next.js)
robosystems-mcp-client	MCP server for AI integration
robosystems-typescript-client	TypeScript SDK
robosystems-python-client	Python SDK
robosystems-content-machine	Open-source content pipeline (video production from SEC data)

Where it's going

A few things on the roadmap that I'd like community input on:

Custom adapters — The adapter architecture is designed to be fork-friendly. Your custom adapters live in adapters/custom_*/ — upstream never touches that namespace. We have SEC active and QuickBooks in development. What data sources would you want to connect?

Subgraph workflows — Subgraphs (workspaces) support AI memory, data forking, and team collaboration today. We're thinking about expanding the workflow patterns — fork parent data, transform in a subgraph, and publish a subset back to the main graph.

How to participate

• Questions → Use the Q & A category
• Architecture discussions → Use the Architecture category
• Bug reports and feature requests → Issues
• Contributing → See CONTRIBUTING.md

If you're trying RoboSystems locally and hit something — a confusing error, a missing doc, a broken command — open an issue. That kind of feedback is the most valuable thing right now.

If you're working with XBRL data, financial knowledge graphs, or AI-powered financial analysis and want to talk architecture, the Architecture category is for that.

Thanks for checking this out.

— Joey