Agent overview
Use Ploid as a model-led research agent with explicit sources, memory, and spend ceilings.
The Ploid Agent accepts a goal in plain language and decides how to complete it within the boundaries you provide. It can search Ploid people data, research the public web, and synthesize the result.
Model-led routing
You describe the outcome, not a fixed internal search sequence. The agent can choose different tools based on the task, evidence already available, and remaining budget.
{
"operation": "ask",
"prompt": "Find ten stage-matched lookalikes for our strongest early-stage customer and cite the evidence for company stage and fit.",
"sources": ["people", "public_web"],
"max_acu": 16,
"memory": "none",
"response_format": "standard"
}Boundaries you control
| Field | Purpose | Default |
|---|---|---|
sources | Allowed source families | people, public_web |
max_acu | Signed admission, billing, and tool-call limit | 2 |
memory | Long-term memory policy | none |
response_format | Standard JSON envelope or Markdown | standard |
output_schema | Optional validated result contract | Omitted |
max_acu accepts whole values from 1 through 64. The gateway signs it into
Eve's authenticated policy; tools enforce the corresponding downstream-call
budget instead of relying on prompt instructions.
Sources
people— Ploid people search, identity, and enrichment capabilities.public_web— public company sites, news, filings, and other web evidence. Omit a source from the list to prevent the agent from using that source family for the turn.
Conversational by default
The normal Agent response is intentionally flexible: data.output contains the synthesis and data.artifacts contains typed tool results. This lets intelligent callers continue reasoning over the answer.
When code needs a fixed contract, opt into structured outputs for that turn only.