Your billing team already asks Claude. Now it can answer.
The denial questions your team types into Claude every day, answered with real payer data instead of a guess. Plus the methodology behind it. All public, all separate from the platform.
What does Upstream open source?
The parts a billing team or a builder can use on their own: a Claude MCP server, Claude Code skills for denial and prior-auth work, open reference implementations of the denial-detection methodology on public data, and a curated list of payer-risk resources. The production engine, the network, and patient data stay private.
Open-source tools
We open the methodology. We keep the engine and your data private.
The open repositories teach the techniques on public CMS data. The production system runs them at scale across the operator network, on live data, behind real controls. No patient information is ever part of what we publish.
Open source, answered plainly.
- Is this the same as the paid platform?
- No. These are public repositories your team can run on their own. The production engine, the network, and patient data stay private and separate from what we publish here.
- Do these tools include patient data or the production model?
- No. The open repositories teach the methodology on public CMS data. It is the technique, not the production weights, and no patient data is ever part of what we publish.
- Do I need an Upstream account to use these?
- No. They are public repositories on GitHub. Install the MCP server or skills directly, no platform account required.
- Which tool should I start with?
- Start with Upstream MCP if your billing team already asks Claude denial questions. Use Upstream Skills for Claude Code workflows, Upstream Community to see the denial-detection methodology, and Awesome Payer Risk for a curated resource list.
Building near this work?
If you are building close to specialty billing, start with the open tools, then talk to us about the platform and the data network behind them.