Tools / Decisions

review_calibration()

Key

Ledger analytics over every scored decision: expected-vs-actual error distributions over time, hit rates by decision type, and repeated variance patterns — the honest measure of whether the workspace's judgment is actually compounding.

review_calibration(window, decision_type)→ CALIBRATION REPORT

Callable over MCP as review_calibration(window, decision_type). REST mirror GET /v1/calibration Planned — alpha is MCP-only

Why your agent calls it

The honest mirror, for the human running the workspace. Expected-vs-actual error over time, hit rates by decision type, and the repeated patterns — like revenue overshooting whenever an assumed elasticity was on the path. If judgment is compounding, this is where it shows.

What it returns

review_calibration() · How good are the calls, over time and by type.ILLUSTRATIVE

Call it from an agent

Over MCP, your agent invokes it on its own when the moment calls for it — just describe the situation. A Python SDK for headless runtimes is planned; this is what the call will look like.

agent · python sdkSDK · PLANNED — ALPHA IS MCP-ONLY

from cascadian import Cascadian

cx = Cascadian(api_key=os.environ["CASCADIAN_API_KEY"])

result = cx.review_calibration(
    window="365d",
    decision_type="pricing",
)

Parameters

Param	Type	Description
window	string	Lookback window, e.g. "365d". Default all time.
decision_type	string	Restrict to one decision type or action.

Example response

review_calibration()JSON · ILLUSTRATIVE

{
  "window": "365d",
  "decisions_scored": 38,
  "hit_rate_within_interval": 0.71,
  "by_action": { "act": 0.74, "wait": 0.81, "test": 0.62 },
  "repeated_variance": [
    "revenue overestimated when discount edges were assumed — 4 of 5 times"
  ]
}

Try it — sandbox

Run this call right here. The response is canned and spec-true — the exact shape your agent gets back over MCP, on sample data.

Sandbox — canned, spec-true responses. No key, no server, no live data.

Sandbox — sample responsereview_calibration()

// Press RUN to call review_calibration() in the sandbox.

Sample — canned from the spec, not a live server

Related — Decisions

keyopen_decision()Open a decision and pin what is knowable now.keyrecord_recommendation()The full recommendation — expectation frozen.scopedrecord_decision()The immutable commit. Approval enforced server-side.keyrecord_outcome()Measured actuals against the frozen expectation.keypropose_learning()Queue what the variance taught — humans approve.keyget_decision()The full decision card, by id.