Tools Reference

PrompTick's MCP server exposes three tools. Each AI Action consumed by these tools counts against your monthly AI Action quota (see Quotas & Limits).

---

generate_prompt

Generate an optimized system + user prompt from a use-case description.

Use when: the user wants to create a brand-new AI prompt template (not when they want to write code).

Input

Field	Type	Required	Description
use_case	string	yes	What the prompt should do (e.g. "customer support chatbot for a SaaS product")
target_model	string	no	Target model ID. Omit to auto-detect.
variables	array	no	Template variables: { name, description, sample_values? }
constraints	string[]	no	Requirements (e.g. "must respond in JSON")
enable_variable_suggestions	boolean	no	Let AI suggest variables. Defaults to true if no variables provided.
skip_clarification	boolean	no	Skip follow-up questions and generate immediately
clarification_answers	array	no	Answers returned from a previous needs_clarification response

Output

On success:

{
  "status": "success",
  "system_prompt": "...",
  "user_prompt": "...",
  "suggested_variables": [{ "name": "...", "description": "..." }],
  "generation_approach": "...",
  "defaults_applied": ["..."]
}

If PrompTick needs more info:

{
  "status": "needs_clarification",
  "message": "...",
  "questions": [{ "questionId": "...", "questionText": "..." }]
}

Re-call with clarification_answers populated.

Example

"Generate a prompt for a bot that reviews pull requests and flags security issues."

---

improve_prompt

Enhance an existing AI prompt.

Use when: the user pastes prompt text and wants it improved.

Input

Field	Type	Required	Description
system_prompt	string	one of	Current system prompt
user_prompt	string	one of	Current user prompt
improvement_instructions	string	no	What to change (e.g. "make it more concise")
target_model	string	no	Target model ID
use_case_context	string	no	What the prompt is used for
variables	array	no	Existing {{variables}} in the prompt — pass them so they're preserved
enable_variable_suggestions	boolean	no	Default false
skip_clarification	boolean	no	Skip clarification
clarification_answers	array	no	Answers to previous clarification questions

At least one of system_prompt or user_prompt is required.

Output

{
  "status": "success",
  "system_prompt": "...",
  "user_prompt": "...",
  "improvement_approach": "...",
  "suggested_variables": [],
  "defaults_applied": ["..."]
}

Or "status": "needs_clarification" with questions (same shape as generate_prompt).

---

enrich_coding_prompt

Turn a coding task description into a rich, well-structured prompt for an AI coding assistant. Two-step tool.

Use when: the user wants to build, create, fix, refactor, or modify code.

Input

Field	Type	Required	Description
coding_request	string	yes	The raw task (e.g. "Add authentication to my Express API")
target_assistant	enum	no	cursor, claude-code, copilot, windsurf, generic (default)
project_context	object	no	{ tech_stack?, description?, language? }
context_answers	array	no	Step 2 only — answers to context requests
task_analysis_passthrough	object	no	Step 2 only — pass back unchanged from step 1

Step 1 — first call

Provide coding_request and project_context. The tool analyses the task and returns context it needs from your codebase:

{
  "status": "needs_context",
  "message": "...",
  "context_requests": [
    { "id": "ctx-001", "type": "file", "description": "..." }
  ],
  "task_analysis_passthrough": {
    "domain": "...",
    "task_summary": "...",
    "...": "..."
  }
}

Step 2 — fulfill context, re-call

Your assistant reads the requested files, then re-calls with:

• same coding_request
• task_analysis_passthrough passed back unchanged
• context_answers, each { requestId, content, fulfilled, note? }

Returns the final enriched prompt:

{
  "status": "success",
  "enriched_prompt": "...",
  "enrichment_summary": "...",
  "target_assistant": "claude-code"
}

Important: don't skip step 1. Letting the tool ask for exactly what it needs produces substantially better prompts than guessing upfront.

---

Errors

Tool errors return isError: true with:

{ "status": "error", "error": "<code>", "message": "..." }

Auth and quota errors come back at the protocol layer — your MCP client typically surfaces them as toast messages. If you hit your monthly AI Action limit, the tool will refuse new calls until reset or upgrade.