Step 5 of 5

Mapping and execution

Define how MCP tool arguments map to upstream request fields, and how the upstream response maps back to MCP output. Then set execution parameters — auth is configured per-upstream in step 2.

Input → Request

yaml / input-mapping
first_name: "$.input.first_name"
last_name:  "$.input.last_name"
email:      "$.input.email"
company:    "$.input.company"
phone:      "$.input.phone"
source:     "$.input.source"

Response → Output

yaml / output-mapping
lead_id:    "$.response.data.id"
status:     "$.response.data.status"
created_at: "$.response.data.created_at"
owner_id:   "$.response.data.owner.id"

Execution

Execution config

Timeouts, retry policy and auth profile reference

Optional

yaml / exec-config
timeout_ms: 10000
retry:
  max_attempts: 3
  backoff_ms: 500
auth:
  profile: acme_bearer
tls:
  verify: true

Live validation and publishing

Live actions save the draft first

Sample uploads, test runs, YAML actions and publish all persist the current draft before calling the backend.

Samples and draft generation

Persist input/output samples, then generate schemas and mappings from real payloads.

Input sample

{
  "first_name": "Ada",
  "last_name": "Lovelace",
  "email": "ada@example.com"
}

Output sample

{
  "id": "lead_123",
  "status": "created"
}

Test run

Run the current draft against the upstream before publishing it.

Test input payload

{
  "first_name": "Ada",
  "last_name": "Lovelace",
  "email": "ada@example.com"
}

Request preview

Response preview

Errors

YAML import and export

Move operation drafts between environments without leaving the wizard.

YAML document

Publish

Save the current draft and expose it to published agents.

Publishing uses the current draft version from this wizard session.

Operation stays editable in this wizard

Saving creates or updates the current draft in place. Use the live actions above to upload samples, test the draft, export or import YAML, and publish when the upstream contract is ready.