Replace the step list atomically

PATCHhttps://myagentmail.com/cadences/{id}/steps

Sends the full ordered list of steps in one call — wipes the previous steps and reinserts. Steps are tiny; diffing is more trouble than it's worth.

Step kinds: • linkedin_connect — POST a connection request (with note) • linkedin_message — DM a 1st-degree connection • email — (v1.1) send via configured inbox • wait — no-op delay before the next step

Conditions (optional, NULL = always run): • after_accept — only run if the previous linkedin_connect step's invite has been accepted (signal.connection_accepted with outbound attribution) • no_reply_to_prev — only run if the lead hasn't replied since the previous step • never_replied — only run if the lead has never replied during this enrollment

Path parameters

id

required

string

Request body

Content-Type: application/json

`steps` required object[]
`kind` required "linkedin_connect" \| "linkedin_message" \| "email" \| "wait"	enum: "linkedin_connect" \| "linkedin_message" \| "email" \| "wait"
`delayDays` required number	Days after the previous step's completion. 0 = fire on enrollment for step 0; 0.25 = 6 hours.
`condition` "after_accept" \| "no_reply_to_prev" \| "never_replied"	enum: "after_accept" \| "no_reply_to_prev" \| "never_replied"
`draftStrategy` "static" \| "webhook"	How the runner gets the rendered content for this step. - `static` — Use `staticBody` (+ `staticSubject` for email) as the message verbatim, with `{{name}}`, `{{first_name}}`, `{{email}}` token substitution. Predictable; zero external deps. - `webhook` — At fire-time, the runner POSTs a `cadence.step.draft.requested` event to the cadence's `webhookUrl` and waits up to 30s for a JSON response of `{ body, subject?, skip?, defer? }`. The customer's app brings the actual content with its own business context, its own LLM if it wants one, its own thread memory. MAM never generates copy on the customer's behalf. The legacy `"ai"` value was REMOVED 2026-05-19 — MAM no longer drafts content server-side. New cadences passing `"ai"` get a `VALIDATION_ERROR` with a migration hint pointing here. Pre-existing rows with `"ai"` will surface `AI_STRATEGY_REMOVED` as the per-enrollment `lastError.code` when the runner tries to fire them — fix by updating the step to `static` (with `staticBody`) or `webhook` (with the cadence's `webhookUrl` set). enum: "static" \| "webhook" default: "static"
`staticSubject` string	Email-only. The literal subject line for static-strategy email steps. Supports `{{name}}` / `{{first_name}}` / `{{email}}` tokens.
`staticBody` string	The literal message body for static-strategy steps. Supports `{{name}}` / `{{first_name}}` / `{{email}}` tokens. Required when `draftStrategy: "static"`.
`promptHint` string	Free-text hint for the step. MAM does not read this — it's passed through verbatim in the `cadence.step.draft.requested` webhook payload (`step.promptHint`) so the customer's draft handler can use it as per-step context. Ignored by static strategy.
`notes` string	Operator notes — not sent anywhere, just stored for the customer's reference.

Responses

200Steps replaced.

Authentication

Send your API key in the X-API-Key header (or Authorization: Bearer <key>). Any prefix accepted by this endpoint — tk_, wk_, ak_, or sa_ — is documented in the key prefix table.

Request

import { MyAgentMail } from "myagentmail";

const client = new MyAgentMail({ apiKey: process.env.MYAGENTMAIL_API_KEY! });

// PATCH /cadences/{id}/steps
const result = await client.request("patch", "/cadences/{id}/steps", { params: { id: "00000000-0000-0000-0000-000000000000" } }, { body: {
  "steps": [
    {
      "kind": "linkedin_connect",
      "delayDays": 0,
      "condition": "after_accept",
      "draftStrategy": "static",
      "staticSubject": "string",
      "staticBody": "string",
      "promptHint": "string",
      "notes": "string"
    }
  ]
} });