Agent quickstart

A compact operating guide for agents helping a user complete the Everyn API quickstart.

Use this page when a coding agent is helping a user complete the API quickstart. The user should be able to point to data, give instructions, approve a limited run, and view results without personally managing every HTTP call.

Help me complete the Everyn API quickstart.

Read:
https://www.geteveryn.com/docs/api/agent-quickstart.md

Start by asking whether I want to use the sample dataset or my own CSV. If you
need an Everyn API key, ask me for it. Run only a limited sample until I approve more.

Agent workflow

Match the Everyn product loop: point to data, give instructions, view results. Keep API mechanics in the background, explain each proposed action in plain language, and ask for approval before starting work that spends credits or expands scope.

Shared rules:

Use https://api.geteveryn.com as the production API base URL.
Use the public OpenAPI schema at https://www.geteveryn.com/openapi/everyn-public-api.json for exact request and response shapes.
If you do not already have an Everyn API key in a secure session context, ask the user for one before making API calls.
Use shell/session values or a real secret manager for credentials.
Do not write API keys to source files, .env files, docs, logs, or committed artifacts.
Never invent credentials.
If an API response includes docsUrl or an error code, open the linked docs and recover before continuing.
Start with a limited sample run. Do not create a larger run until the user explicitly approves it.
Summarize created resource IDs so the user can return to the dataset, job spec, run, or export later.

Sample dataset path

Ask whether the user wants to use the sample dataset or their own CSV.
If the user chooses the sample, explain the sample leads.csv in the quickstart and propose a practical outcome for the run.
Let the user approve that outcome or provide their own instructions.
Draft the job spec and show the user the sample-run plan.
Ask for approval before creating and starting the limited sample run.
Communicate status while the upload, run, and inspection steps progress.
When the sample run is complete, inspect outputs, failures, events, and usage.
Export the reviewed sample results, analyze the CSV for the user, and summarize created resource IDs.

Custom CSV path

Ask the user for the CSV path or location.
Sample and profile the CSV before uploading it: summarize columns, row count if practical, representative rows, missing values, and any obvious data-quality concerns.
Propose one to three Everyn run outcomes that look possible for the dataset.
Let the user approve one outcome or provide their own instructions.
Draft the job spec and show the user the sample-run plan.
Ask for approval before uploading the dataset and starting the limited sample run.
Communicate status while the upload, run, and inspection steps progress.
When the sample run is complete, inspect outputs, failures, events, and usage.
Export the reviewed sample results, analyze the CSV for the user, and summarize created resource IDs.

Technical exit conditions

Stage	Exit condition
Upload	The dataset upload reaches `completed` and returns `datasetId`.
Job spec	The reusable job spec is created with a validated `outputSchema`.
Sample run	A limited run reaches `succeeded` or `completed_with_flags`.
Review	Outputs match the schema and failures are understandable.
Export	A reviewed run creates a `ready` export and downloads CSV bytes.

Recovery

Use Authentication, Idempotency, API errors, Dataset uploads, and the OpenAPI reference when the quickstart path needs more detail.