# Pull past performance for a proposal

To pull past performance with an AI agent, ask it to call search_awards with your recipient_uei and an action_from date, then rank the awards against the target solicitation's NAICS, agency and scope. It returns PIID, agency, obligated and potential value, dates and description for each, the facts a past performance reference needs.

Proposal writers assembling a past performance volume against a deadline. Evaluators check recency and relevance first, and both come straight from award data: the period of performance and how close the scope and agency are to the new requirement.

## The prompt

```
From the federal awards for UEI ABCDEFGH1234 since 2021-10-01, pick the three most relevant past performance references for a VA IT modernization bid under NAICS 541512. For each, give PIID, agency, value, period of performance and a one-line scope.
```

## What the agent calls

1. search_awards: The candidate pool; the agent ranks by NAICS match, agency match, value and recency.

```json
{"name":"search_awards","arguments":{"recipient_uei":"ABCDEFGH1234","action_from":"2021-10-01","limit":100,"user_request":"past performance references for VA bid"}}
```

## Sample output

Trimmed; field names and structure match the live response, values are shortened or illustrative.

```json
{
  "data": [
    {
      "piid": "36C10B24C0042",
      "awarding_agency": "Department of Veterans Affairs",
      "naics": "541512",
      "description": "EHR INTERFACE MODERNIZATION SUPPORT",
      "start_date": "2024-03-01",
      "potential_end_date": "2027-02-28",
      "obligated": 3100000,
      "potential_value": 5400000
    }
  ]
}
```

## How to read the answer

- Most solicitations define recency (for example within the last three or five years). The agent should drop awards outside that window before ranking.
- Relevance usually weighs scope, size and complexity. Value and period come from the data; scope similarity is the agent's judgement and should be marked as such.

## Make it yours

- Ask the agent to fill your past performance template with the chosen awards.
- Include subcontract work you performed yourself; it is not in prime award data.

## Pitfalls

- Contracting officer references and CPARS ratings are not in USAspending. Add them from your own records.
- Never let the agent describe work you did not do; descriptions in award data are short and generic.

## The same job over REST

```bash
curl "https://samgovapi.io/api/v1/awards?recipient_uei=ABCDEFGH1234&action_from=2021-10-01&limit=100" -H "Authorization: Bearer $SAMGOVAPI_API_KEY"
```

Plan: every plan. Credits: 1 credit per 100 rows returned, minimum 1.

## Questions

### Why not just use the entity award history endpoint?

GET /entities/{uei}/awards is the history endpoint at 3x credits. search_awards with recipient_uei and a date filter usually answers this job for less.

### Are task orders included?

Yes. Orders under IDIQs and BPAs appear with their parent_piid, which is useful to show the vehicle.

All recipes: https://samgovapi.io/agents/recipes.md

Independent. Not affiliated with, endorsed by or sponsored by GSA, SAM.gov or any U.S. government agency. Registering in SAM.gov is free at sam.gov.