> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nimbleway.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Enrichment
> Send rows with gaps, get them back complete — every added value cited
**Enrichment** agents complete the data you already have. Send partial records in `input_data` — CRM leads missing funding stages, product rows missing specs — and the agent researches each gap on the live web and fills it. Your values are treated as ground truth and passed through untouched. Only the blanks get researched.
```text theme={"system"}
IN "company_name": "Acme Corp"
"headquarters": null
"key_contacts": null
── the agent researches only what's missing ──
OUT "company_name": "Acme Corp" · yours, untouched
"headquarters": "Shelton, CT" · cited: acme.com
"key_contacts": [{"name": "..."}] · cited: acme.com/leadership
```
## Example Request
An enrichment run needs the rows (`input_data`) and the shape (`output_schema`, on the request or the agent). The `lead-enrichment` template ships a schema covering company identity, headquarters, decision-makers, and buying signals:
```python Python theme={"system"}
agent = nimble.agents.create(template="lead-enrichment")
run = nimble.agents.runs.create(
agent.id,
input="Fill in the missing fields for these companies.",
input_data=[
{"company_name": "Nimble", "headquarters": None, "key_contacts": None},
{"company_name": "Bright Data", "headquarters": None, "key_contacts": None},
],
)
```
```typescript TypeScript theme={"system"}
const agent = await nimble.agents.create({ template: "lead-enrichment" });
const run = await nimble.agents.runs.create(agent.id, {
input: "Fill in the missing fields for these companies.",
input_data: [
{ company_name: "Nimble", headquarters: null, key_contacts: null },
{ company_name: "Bright Data", headquarters: null, key_contacts: null },
],
});
```
```go Go theme={"system"}
agent, err := client.Agents.New(ctx, nimble.AgentNewParams{
Template: param.NewOpt("lead-enrichment"),
})
run, err := client.Agents.Runs.New(ctx, agent.ID, nimble.AgentRunNewParams{
Input: "Fill in the missing fields for these companies.",
InputData: nimble.AgentRunNewParamsInputDataUnion{
OfMapOfAnyMap: []map[string]any{
{"company_name": "Nimble", "headquarters": nil, "key_contacts": nil},
{"company_name": "Bright Data", "headquarters": nil, "key_contacts": nil},
},
},
})
```
```bash CLI theme={"system"}
nimble agents:runs create \
--agent-id "$AGENT_ID" \
--input "Fill in the missing fields for these companies." \
--input-data '[{company_name: "Nimble", headquarters: null, key_contacts: null}, {company_name: "Bright Data", headquarters: null, key_contacts: null}]'
```
```bash cURL theme={"system"}
curl -X POST "https://sdk.nimbleway.com/v2/agents/$AGENT_ID/runs" \
--header 'Authorization: Bearer ' \
--header 'Content-Type: application/json' \
--data '{
"input": "Fill in the missing fields for these companies.",
"input_data": [
{"company_name": "Nimble", "headquarters": null, "key_contacts": null},
{"company_name": "Bright Data", "headquarters": null, "key_contacts": null}
]
}'
```
`input_data` accepts a list of partial rows (or a single object) mirroring the output schema. An `output_schema` — on the request or the agent — is required.
## Example Response
The result is `type: "json"` — your rows back, gaps filled:
```json theme={"system"}
{
"type": "json",
"content": [
{
"company_name": "Nimble",
"website": "https://www.nimbleway.com",
"headquarters": "New York, NY, United States",
"key_contacts": [
{ "name": "Uri Knorovich", "title": "Co-founder & CEO", "seniority": "c_suite" }
],
"buying_signals": ["Launched the Web Search Agents API (July 2026)"],
"summary": "Nimble is an AI-native web data platform providing search, extraction, and research agent APIs."
}
],
"trust": {
"confidence": "high",
"reasoning": "High confidence - cells fill rate is 100% and trust per claim ratio is 94%",
"claims": [
{
"path": "$[0].company_name",
"confidence": "pre_existing",
"reasoning": "Provided in input_data"
},
{
"path": "$[0].headquarters",
"confidence": "high",
"reasoning": "Backed by a primary source (official)",
"citations": [
{
"url": "https://www.nimbleway.com/about",
"excerpts": ["Headquartered in New York"],
"source_category": "official"
}
]
}
]
}
}
```
## Trust in enrichment runs
Every value in the result carries a claim per JSON path in the [trust report](/nimble-sdk/web-search-agents/trust). The grades distinguish your data from researched data:
| Confidence | Meaning in an enrichment run |
| ------------------------- | ---------------------------------------------------- |
| `pre_existing` | Came from your `input_data` — carried over unchanged |
| `high` / `medium` / `low` | Researched — with citations and excerpts |
For example, `$[0].headquarters` carries a citation and a `high` grade, while `$[0].company_name` is marked `pre_existing`. Route low-confidence fills to review and sync the rest into your CRM, with source URLs alongside.
Two additional rules:
* The run's overall grade counts only researched values. `pre_existing` data is passed through, never graded.
* A gap the agent could not fill comes back as a cited `null` ([verified absence](/nimble-sdk/web-search-agents/trust#trust-for-structured-outputs)), distinguishing "no funding data exists" from "not found".
## Use cases
* **CRM hygiene** — complete funding, headcount, and tech-stack fields across the pipeline, with buying signals like recent hiring flagged (`lead-enrichment`).
* **Financial datasets** — fill valuation, investor, and filing gaps from primary sources (`financial-intelligence`).
* **Product catalogs** — complete specs, pricing, and availability across competitor listings (`ecommerce-intelligence`).
To enrich thousands of rows on a schedule, hand the same agent to [Jobs](/nimble-sdk/agentic/jobs) — CSV or Parquet in, enriched rows delivered to your warehouse.
How per-path claims, citations, and `pre_existing` fit together