A general-purpose LLM hallucinates a Travel CRNA rate. ClinicalRate gives your agent the live distribution, the sample size, the as-of timestamp, and the canonical role ID — every time.
A general LLM's answer to a rate question is a weighted average of forum posts and training cutoff. Useful for context. Useless for a quote.
You don't want your workforce agent making up numbers. You want it citing the live 50th percentile, with sample size and recency, sourced from one canonical taxonomy.
A single endpoint accepts the user's question and returns the canonical role ID, the live p50/p75, sample size, as-of timestamp, and source count — the exact citations your agent needs to be defensible.
CR-0091.$ curl https://api.clinicalrate.com/v1/grounded \ -H "Authorization: Bearer $CR_TOKEN" \ -d '{"role":"CRNA","geo":"HOU","scope":"travel"}' { "canonical_role": "CR-0091", "label": "Travel CRNA", "geo": "Houston-Sugar Land-Baytown", "bill_rate": { "p50": 291.00, "p75": 312.00, "delta_wow": 0.062, "n": 312 }, "as_of": "2026-05-11T07:00:00Z", "sources": 6, "winsor": "1.5x_IQR" }
POST /v1/grounded · median latency 92ms · OAuth 2.0 · rate-limit 1,000 rpm
Index the canonical taxonomy + percentile snapshots in your vector store, refreshed daily. The agent retrieves the live distribution and cites it inline. No model fine-tuning required.
Drop the ClinicalRate tool spec into your agent's tool list. get_rate(role, geography, percentile) returns a structured payload with provenance — ready for your agent to render.
License a frozen historical snapshot as the ground-truth set for your rate-recommendation evals. Know whether your agent is getting better or worse over time, against a real-market reference.
# Tool definition (OpenAPI 3.1 fragment) { "name": "get_clinicalrate_rate", "description": "Return live bill-rate distribution for a clinical role × geography.", "parameters": { "role": "string — canonical role ID or natural-language title", "geography": "string — state, metro, or 'national'", "as_of": "date — optional, defaults to today" }, "returns": { "role_id": "CR-XXXX (canonical)", "percentiles": "{ p25, p50, p75, p90 }", "sample_size": "int", "as_of": "ISO 8601 timestamp", "trend_30d_pct": "float", "sources": "array — provenance per ingestion partner" } }
| Grant | Inference license | Training license |
|---|---|---|
| Real-time API calls into a deployed agent | ✓ included | ✓ included |
| Cache responses for up to 24 hours | ✓ included | ✓ included |
| Index canonical taxonomy + snapshots in a vector store | ✓ included | ✓ included |
| Use as fine-tuning data for proprietary models | — | ✓ included |
| Use as ground-truth for evaluation suites | ✓ included | ✓ included |
| Re-distribute snapshots to third parties | — | — |
“The eval surface alone justified the subscription. Before ClinicalRate we had no way to score whether our agent's rate suggestions were getting closer to the market or further from it.”
Yes, under the training license. The frozen monthly snapshot is licensable as fine-tuning data for proprietary models. Redistribution remains restricted.
Yes. The MCP server is included in the ClinicalRate plan (currently in private preview). It exposes the read API as MCP tool calls with full provenance in the response envelope.
One plan: $2,500 a month, billed monthly, cancel anytime — API and MCP access included. If you need fine-tuning rights or very high call volumes, talk to us.
Book a 30-minute call with our data team. Bring your agent — we'll wire up the tool call live in the meeting.
Routed directly to aaron@clinicalrate.com — no SDR queue.