For AI agents
Specialized risk intelligence, one call away
Precise L1/L2/L3 factor decomposition and ETF hedge ratios for ~3,000 US equities — JSON-native, metered, and reproducible. An agent calls it in one shot instead of reasoning from training data.
Approving us for your agent fleet? This page is the one-pager for risk & compliance teams — discovery artifacts, data policy, reproducibility guarantees, and pricing are all linked below and published machine-readably.
Why call us instead of reasoning
A general model approximating a factor decomposition from training data is slow, expensive, and unauditable. A single RiskModels call returns the exact, current four-layer attribution and the ETF hedge map — with a model version and data date you can cite. Live processing latency and service health are published at /api/health.
Discover us programmatically
Four machine-readable entry points. Start with the manifest — it links to the rest.
Capability manifest
One fetch: capabilities, pricing, reproducibility, and cross-links to every other artifact.
/.well-known/agent-manifest.json →OpenAPI spec
Full REST contract with per-endpoint x-pricing extensions.
/openapi.json →MCP server
Streamable-HTTP MCP endpoint with discovery + live-data tools.
/.well-known/mcp.json →llms.txt
Crawler-friendly summary, including a shared demo key when the host enables one.
/llms.txt →Onboard without a human
An agent goes from never-heard-of-us to first paid call in three requests. Bearer auth (rm_agent_* / rm_user_*), idempotent retries, and metered pricing. Full key management lives at get a key and pricing.
# 1. Self-provision free starter credits (no human in the loop)
curl -X POST https://riskmodels.app/api/auth/provision-free
# 2. Call any endpoint with the Bearer token you receive
curl -X POST https://riskmodels.app/api/decompose \
-H "Authorization: Bearer rm_agent_..." \
-H "Content-Type: application/json" \
-d '{"ticker":"NVDA"}'
# 3. Retry safely — same Idempotency-Key never double-charges
# -H "Idempotency-Key: <uuid>"Generate tool definitions for your framework:
# Emit ready-to-paste tool definitions for your agent framework
npm install -g riskmodels@latest
riskmodels manifest --format anthropic # or: openai | zedTrust & compliance signals
Built for regulated workflows — 13F structure, fund attribution, portfolio risk.
Point-in-time & reproducible
Decompositions are deterministic and replayable back to 2006-01-04. Every response carries model_version, data_as_of, and factor_set_id so any number can be cited and re-derived.
Auditable & disclosed
request_id correlates response → billing event. Data handling, retention, and training policy are published machine-readably at /.well-known/agentic-disclosure.json.
Cost in the protocol
Every call returns _agent.cost_usd in the body and X-API-Cost-USD in headers. Pay-per-call, $20 free credits, no subscription. HTTP 402 when balance runs low — never a surprise.