Agent quickstart

Create a project, add context, and start the first run.

External agents should begin with a known project, grounded seller context, and a queued discovery run. After that, read claims and next actions before advancing the journey.

Project

Create the Vera workspace.

Use one project per outbound motion. The project name, business name, and website seed the seller context that later runs will rely on.

Create project

$ curl -X POST https://api-vera.symph.ai/v1/projects \
>   -H "Authorization: Bearer vwk_live_..." \
>   -H "Content-Type: application/json" \
>   -d '{
>     "name": "First outbound campaign",
>     "business_name": "Vera.AI",
>     "website": "https://vera.symph.ai"
>   }'

Context

Upload the first seller context.

Add notes, URLs, and files before prospecting. Vera turns this source material into knowledge claims with status, trust, and evidence metadata.

Add context

$ curl -X POST https://api-vera.symph.ai/v1/projects/proj_01hx6x9v2r/context \
>   -H "Authorization: Bearer vwk_live_..." \
>   -F "notes=ICP, positioning, proof, objections, tone rules" \
>   -F "urls=https://vera.symph.ai" \
>   -F "files=@./product-one-pager.pdf"

First run

Queue business discovery.

Discovery normalizes raw context into reviewable seller knowledge. Poll the run, then read knowledge and next actions before queueing prospect work.

Discovery run

$ curl -X POST https://api-vera.symph.ai/v1/projects/proj_01hx6x9v2r/runs \
>   -H "Authorization: Bearer vwk_live_..." \
>   -H "Content-Type: application/json" \
>   -d '{
>     "operation": "discover_business",
>     "inputs": { "refresh": false }
>   }'