Documentation

RiskState is a deterministic engine that converts live market state into dynamic risk permissions before execution. Use it from any language over HTTP, from an AI agent via the MCP server, or as a reference for risk governance design.

Integrate

API Reference (v1) → The POST /v1/risk-state endpoint. Auth, request and response schemas, action enums, risk flags, and pre-trade workflows for spot, perpetual futures, DeFi borrowing, and AI agents. Production-stable.

API Reference (v2 alpha) → POST /v2/portfolio-risk-state (multi-position aggregation, correlation, BTC-equivalent net, stablecoin breakdown), POST /v2/stress-scenario (6 deterministic tail templates with baseline-aware diff), and POST /v2/backtest/shadow (multi-asset shadow backtest with per-snapshot policy assessment and aggregate distribution metrics — the evidence engine for pilots). Pre-stable — alpha.5 / alpha.2 / alpha.2.

MCP Server → The @riskstate/mcp-server npm package. A single get_risk_policy tool that lets any MCP-compatible agent (Claude Desktop, IDE extensions, custom MCP clients) consume RiskState directly.

Visualizer Guide → How to read every section of the free RiskState Visualizer: ticker bar, Market State grid, Risk Permissions, composite, structural and tactical scores, policy hash verification.

Operations

Institutional Integration Guide → Pre-vendor due diligence brief. Architecture, data classification, auth, audit, retention, incident response, shadow-mode pilot model, roadmap to production-ready, and SIG Lite mapping. Read this first if you are evaluating RiskState for a fund, desk or treasury team.

Audit Log → Every authenticated decision from /v1/risk-state, /v2/portfolio-risk-state, /v2/stress-scenario and /v2/backtest/shadow is persisted with policy_hash (where applicable), scoring_version, request shape, response summary and latency_ms. Read via GET /v2/audit/decisions with endpoint discriminator per record. 365-day retention. SIG Lite procurement mapping included.

Policy Config → Per-API-key risk profile that further tightens the engine's caps. Tighten only — never loosens. Configure max_gross, max_per_asset, leverage_cap, blocked_protocols per customer mandate via PUT /v2/keys/policy. Enforced on /v1/risk-state today — v2 endpoints record the active profile in audit but cap-enforcement plumbing lands jointly with v2 stabilisation at the freeze lift.

The Engine

Methodology → Public methodology of the engine: the 4 caps, the 5-level policy engine, composite / structural / tactical scoring, the policy combiner, market regimes, and the policy_hash audit contract.

Backtest → The framework used to validate scoring: forward-return analysis, permutation tests, circular block bootstrap, Sharpe per quintile, effective sample size, reproducibility discipline.

Research → Dated, citable technical research. Papers on a six-month cadence, technical posts monthly. Framework findings and out-of-sample reports.

Get an API key

Request API access from the home page — only an email is required. Keys are delivered by email instantly and are free during beta.

Support

Questions, bug reports, and feedback: hello@likido.xyz.