A non-custodial Risk, Strategy & Evidence OS for programmable assets.
Ryntra turns an intended economic action into checkable evidence: what was proposed, what the evidence supported, what a human authorized, and what the network actually did — kept as four separate facts rather than one confident summary.
| Repository | What it is |
|---|---|
| ryntra-agent-control | Read-only SDK, local MCP server, OpenAPI contract and a runnable reference flow for Agent Control. Apache-2.0. |
| ryntra-arc-demo | Arc Public Testnet reference implementation for the decision and settlement evidence model. MIT. |
These are deliberate, bounded surfaces: types, contracts, interfaces, examples and tests that a developer can read and run. The production application, its infrastructure, the risk logic and internal planning are maintained privately. Nothing here is a hosted service, and no package is published to a registry.
Ryntra holds no keys, signs nothing, and moves no funds. Reads are bounded to fixed provider origins and fail closed rather than guessing. A passing integrity check proves that a document is internally consistent — never that an action was authorized, executed or settled.
Built by Dmytro Plavutskyi · ryntra.io