Tools / Causal

identify()

Key

Runs the identification strategy tree on the approved subgraph: randomized → backdoor (pgmpy adjustment sets) → frontdoor → IV if declared → DiD/ITS if preconditions are declared → not_identifiable / exploratory_only. Returns the full review, including forbidden controls and every required assumption, each linked to supporting or missing memory. No estimate exists without this review — invariant I3.

identify(treatment, outcome, as_of)→ IDENTIFICATION REVIEW

Callable over MCP as identify(treatment, outcome, as_of). REST mirror POST /v1/identify Planned — alpha is MCP-only

Why your agent calls it

The question almost nobody asks before quoting an effect: can this even be estimated from data we have? identify() answers it on the approved graph — here, discount → win_rate is identifiable by backdoor with adjustment set {segment, rep_experience} — and names the forbidden controls and the assumptions doing the work. Without its review id, estimate() refuses to run.

What it returns

identify() · Can this effect be estimated at all? The gate.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.identify(
    treatment="discount",
    outcome="win_rate",
    as_of="2026-06-11T17:00:00Z",
)

Parameters

Param	Type	Description
treatmentREQUIRED	string	The cause variable, e.g. "discount".
outcomeREQUIRED	string	The effect variable, e.g. "win_rate".
as_of	timestamp	Identify on the graph as it stood then. Default now.

Example response

identify()JSON · ILLUSTRATIVE

{
  "query": { "treatment": "discount", "outcome": "win_rate" },
  "strategy": "backdoor",
  "identifiable": true,
  "adjustment_set": ["segment", "rep_experience"],
  "forbidden_controls": ["deal_size_post_quote"],
  "required_assumptions": [
    { "assumption": "no unobserved confounder of discount and win_rate",
      "support": "mem_0871" }
  ],
  "identification_review_id": "idr_0271"
}

Related — Causal

keypropose_nodes()Propose variables for the graph — status proposed.keypropose_edge()Propose a cause→effect edge, with evidence.keyget_graph()The governed graph, by version or as-of.keyvalidate_graph()Structural checks before anyone trusts an edge.keyestimate()A causal estimate that survives the skeptic checks.keysimulate_intervention()do(X=x) on the approved graph — the knock-on report.keyattribute_change()What caused this metric to move?