---
title: "Letter of recommendation"
method: POST
path: "/v1/tools/rec"
---

# Letter of recommendation

`POST /v1/tools/rec`

## Request

- Base URL: `https://api.counselorai.app`
- URL: `https://api.counselorai.app/v1/tools/rec`
- Auth: HTTP bearer

Draft a letter of recommendation from counselor notes. Call this to WRITE a letter. To critique one that already exists, call letterlens instead, which is much cheaper. Runs the same 5-step pipeline as appeal.

## Headers

- `Idempotency-Key` string
- `X-Max-Credits` integer

## Request body

- object
  - `studentName` string: The student this is about. Used in the output text.
  - `counselorName` string: Who the letter is from. The counselor stays the author; this tool drafts, it does not sign.
  - `purpose` 'college_general' | 'college_honors' | 'scholarship' | 'transfer' | 'professional' | 'employment' | 'program' | 'undergraduate' | 'graduate' | 'internship' | 'other': What the letter is for. undergraduate, graduate, internship and other are legacy aliases kept working for existing integrations.
  - `profType` 'law' | 'medical' | 'mba' | 'dental' | 'pharmacy' | 'pa' | 'nursing' | 'socialwork': Only when purpose is professional: which professional school, since each reads for different things.
  - `recipient` string: Who receives it, if named.
  - `schools` string: Target school or program.
  - `major` string
  - `selectivity` string: How selective the target is. Shifts how much evidence the letter has to carry.
  - `relationship` string: How the counselor knows the student, and for how long.
  - `traits` string, required: Three standout traits. Send prose, not a list of adjectives: the pipeline turns each one into a scene and needs something to work from.
  - `story` string: A specific moment worth telling. The single highest-value field here.
  - `contribution` string: What the student adds to a campus.
  - `resumeText` string: Resume or activity list as plain text.
  - `sampleText` string: A sample of the counselor's own writing, so the letter lands in their voice rather than a generic one.

## Response `200`

Tool output

- object
  - `request_id` string
  - `tool` string
  - `output` object
    - `text` string, required: The generated output.
    - `meta` object: Pipeline metadata (pipeline, verify status, fix pass).
    - `verification` object: Whether a model verify pass checked this text, and its verdict.
      - `checked` boolean
      - `method` string, nullable
      - `status` 'PASS' | 'FAIL' | 'null', nullable
      - `score` integer, nullable
      - `fix_applied` boolean, nullable
    - `grounding` object: What the generation was anchored to. fact_sheet is the exact ground truth the verify pass checked dollar figures, names, and dates against; null for tools without a fact-grounded verify pass.
      - `fact_sheet` string, nullable
      - `fields_reflected_in_prompt` string[]
      - `fields_mapped_or_transformed` string[]
      - `fields_not_reflected` string[]
    - `input_fidelity` string: 'strict': this lane never truncates silently; an over-limit input is rejected with a validation error, so the output was built on exactly what was sent.
  - `billing` object
    - `status` 'settled' | 'pending'
    - `credits_charged` integer, nullable
    - `credits_remaining` integer, nullable

## Other responses

- `402`: Insufficient credits or approved spending limit exceeded. Inspect error.code.
- `429`: Rate limited. Retry after the Retry-After header.

---

[API](https://skmtc.dev/counselorai/apis/counselorai-api.md) · [All operations](https://skmtc.dev/counselorai/apis/counselorai-api/llms.txt) · [OpenAPI document](https://skmtc.dev/counselorai/apis/counselorai-api/revisions/1d93eed35463?raw)