HyperPrompt API reference

POST/api/enhance

curl -X POST https://hyperpromptai.com/api/enhance \
  -H "Authorization: Bearer $HYPERPROMPT_API_KEY" \
  -H "content-type: application/json" \
  -d '{"type":"prompt","content":"write a launch email for our analytics feature","mode":"improve"}'

{
  "success": true,
  "data": {
    "enhanced": "<enhanced prompt text>",
    "tokens": 512,
    "originalLength": 47,
    "enhancedLength": 940
  }
}

HyperPrompt picks the model; it cannot be set per request. Endpoint errors: 400 “Tool is required for rule enhancement”, 429 “Too many enhancement requests; try again in a minute” (retry-after: 60), 502 with code: "AI_PROVIDER_UNAVAILABLE" (retry-after: 5), 503 “AI service is not configured”.

GET/api/library?type=prompt

type is optional: prompt, rule, skill, or mcp. Without it, all four are returned. Results are sorted newest-updated first and are not paginated. Prompts include your personal prompts, prompts in workspaces you belong to, and public prompts you saved.

[
  {
    "id": "<prompt id>",
    "type": "prompt",
    "title": "Release notes writer",
    "description": "",
    "content": "<prompt text>",
    "href": "/prompts/general/release-notes-writer",
    "isPublic": false,
    "isFavorite": false,
    "isDraft": false,
    "currentVersion": 3,
    "variables": ["changes"],
    "folder": "Marketing",
    "workspaceName": "Acme team",
    "approvalStatus": "approved",
    "createdAt": "2026-11-01T10:00:00.000Z",
    "updatedAt": "2026-11-04T09:30:00.000Z"
  }
]

Optional keys are omitted when empty. Rules add tool; skills add version. Errors use {"error": "..."}: 400 “Invalid library type”, 500 “Failed to fetch library items”.

POST/api/prompts

curl -X POST https://hyperpromptai.com/api/prompts \
  -H "Authorization: Bearer $HYPERPROMPT_API_KEY" \
  -H "content-type: application/json" \
  -d '{"title":"Release notes writer","content":"Summarize these changes for customers: {{changes}}","variables":["changes"],"isPublic":false}'

{
  "id": "<prompt id>",
  "title": "Release notes writer",
  "slug": "release-notes-writer",
  "category": "general",
  "isPublic": false,
  "isDraft": false,
  "currentVersion": 1,
  "approvalStatus": "approved",
  "workspaceId": null,
  "createdAt": "2026-11-05T12:00:00.000Z"
}

Version 1 is created automatically with the note “Initial version”. Errors: 404 “Workspace not found” (no edit access to workspaceId), 429 “Too many submissions; try again later”.

GET/api/prompts/{id}

Returns {"success": true, "data": { ...prompt, "author": {...}, "workspace": {...}, "similarPrompts": [...] }}. You can read your own personal prompts, prompts in your workspaces, and any public prompt.

PATCH/api/prompts/{id}

Send at least one field. Accepts the create fields above plus folderId (your folder, or null), workspaceId (or null to make it personal), approvalStatus (draft, in_review, or approved; only workspace owners can approve), and changeNote (up to 300 characters). Unknown fields are rejected. Every successful update creates a new version.

curl -X PATCH https://hyperpromptai.com/api/prompts/<prompt id> \
  -H "Authorization: Bearer $HYPERPROMPT_API_KEY" \
  -H "content-type: application/json" \
  -d '{"content":"Summarize these changes for existing customers: {{changes}}","changeNote":"Narrow audience"}'

Response: {"success": true, "data": { ...prompt }, "isPublic": false}; data.currentVersion is the new version number. Errors: 404 “Prompt not found” (also returned when you lack edit access), 400 “Folder not found” or “Workspace not found”, 403 “Only the workspace owner can approve prompts”.

DELETE/api/prompts/{id}

Permanently deletes the prompt and its versions. Personal prompts can be deleted by their author; workspace prompts by owners and editors. Returns {"success": true}.

GET/api/prompts/{id}/versions

Returns {"success": true, "data": [ ...versions ]}, newest first, up to 100. Each version has version, title, content, description, category, the list fields, isPublic, isDraft, approvalStatus, changeNote, and createdAt.

POST/api/prompts/{id}/versions

Restores an earlier version as a new version. Body: {"version": 2, "changeNote": "optional"}. The default note is “Restored version 2”. Restoring an approved workspace prompt sets it back to in_review. Errors: 404 “Version not found” or “Prompt not found”.

POST/api/prompts/import

Body: {"prompts": [ ... ]} with 1–100 prompt objects using the create fields title, content, description, category, and the list fields. All prompts are created as private drafts in one transaction, so either all are imported or none are. Returns 201 with {"success": true, "data": [{"id": "...", "title": "..."}], "count": 1}. See Import prompts from JSON.

GET/api/prompts/{id}/comments

Returns up to 200 comments, oldest first, each with body, createdAt, and user (id, name, image).

POST/api/prompts/{id}/comments

Body: {"body": "1–2,000 characters"}. Works on public prompts, your own prompts, and prompts in any workspace you belong to (all roles). On workspace prompts, the comment is added to the activity feed and notifies other members. Returns 201.

DELETE/api/prompts/{id}/comments/{commentId}

Allowed for the comment’s author, the prompt’s author, or the workspace owner. Otherwise 403 “Forbidden”.

GET/api/evaluations

Lists your suites with case and run counts and the latest run summary.

POST/api/evaluations

Body: name (required, up to 120), promptTemplate (required, up to 40,000; use {{input}}), description (up to 1,000), rubric (up to 4,000). Returns 201 with the suite.

GET · PATCH · DELETE/api/evaluations/{id}

GET returns the suite with its cases and 10 most recent runs including results. PATCH updates name, description, promptTemplate, or rubric. DELETE removes the suite.

POST/api/evaluations/{id}/cases

{
  "name": "Mentions pricing",
  "input": "Draft a launch email for the analytics feature",
  "expectedOutput": "analytics dashboard pricing",
  "variables": { "tone": "friendly" },
  "assertions": {
    "requiredTerms": ["pricing"],
    "forbiddenTerms": ["guarantee"],
    "minLength": 200,
    "maxLength": 1500
  }
}

Only name is required. Up to 20 cases per suite (409 “Each suite can have up to 20 cases”). Delete a case with DELETE /api/evaluations/{id}/cases/{caseId}.

POST/api/evaluations/{id}/run

Body: {"models": ["provider/model"], "judgeModel": "provider/model"}. Use 1–3 OpenRouter model IDs; judgeModel is optional. Requires Pro or Team. The response returns when every call has finished: {"success": true, "data": { "id": "...", "status": "completed", "results": [ ... ] }}. Each result has caseId, model, output, matchScore, assertionScore, assertionDetails, judgeScore, judgeReason, latencyMs, tokens, and error.

Errors: 403 “Multi-model evaluations require a paid plan”, 409 “Add at least one test case”, 400 “A run can make at most 10 output and 10 judge requests”, 429 “Too many AI requests; try again in a minute”, 500 “Evaluation run failed”.

GET/api/workspaces

Lists workspaces you belong to, each with your role and member and prompt counts.

PATCH/api/workspaces/{id}

Owner only. Body: {"name": "New name"} (1–100 characters). Membership, invitations, and ownership are covered in Team workspaces.

GET/api/account/export

Returns {"exportedAt": "<ISO timestamp>", "account": { ... }} with content-disposition: attachment; filename=hyperprompt-export.json. Requires Pro or Team (403 “Data export requires Pro”). The contents are listed in Export your account.

curl https://hyperpromptai.com/api/account/export \
  -H "Authorization: Bearer $HYPERPROMPT_API_KEY" \
  -o hyperprompt-export.json