prompt-improver-mcp

Improve a prompt

improve_prompt

Rewrites rough or follow-up prompts into self-contained, model-ready prompts by applying research-backed prompting techniques, returning citations, assumptions, and clarifying questions.

Instructions

Rewrite a prompt so a model answers it better, using techniques from the prompting literature: LLaVA-CoT staged reasoning, System 2 Attention context distillation, conversational query rewriting, Rephrase-and-Respond, Chain-of-Verification, Step-Back, Plan-and-Solve, Least-to-Most, and more. The server cannot see the conversation, so pass the current prompt verbatim plus the earlier turns and context it depends on. Returns the improved prompt, the techniques used (with paper citations), what changed, assumptions, and clarifying questions. If backend_used is 'host', improved_prompt is a template draft: finish it by following host_instructions.

Input Schema

TableJSON Schema

Name	Required	Description	Default
`prompt`	Yes	The user's current prompt, verbatim.
`target`	No	'reasoning' for models with built-in thinking (Claude, o-series, R1): no visible chain-of-thought cues. 'standard' for models without it (e.g. small open VLMs): adds explicit step-by-step and LLaVA-CoT stage tags.	reasoning
`backend`	No	'auto' uses the Claude API if credentials are available, else MCP sampling if the client supports it, else 'host' (a template draft for the calling model to finish).	auto
`context`	No	Other material the prompt depends on: files, documents, task background.
`history`	No	Earlier conversation turns, oldest first, including previous assistant responses.
`has_image`	No	True if the prompt will be sent with an image or video.
`techniques`	No	Force exactly these techniques instead of automatic selection. Ids: context_distillation, self_contained_rewrite, failure_feedback, rephrase_expand, clarify, visual_grounding, step_back, plan_and_solve, least_to_most, staged_reasoning, verification, zero_shot_cot, re_reading, specificity_format
`refine_rounds`	No	Self-Refine critique-and-revise rounds (Claude API backend only).

Output Schema

TableJSON Schema

Name	Required	Description
`notes`	No
`target`	Yes
`changes`	Yes
`profile`	Yes
`techniques`	Yes
`assumptions`	Yes
`backend_used`	Yes
`inference_tips`	Yes
`improved_prompt`	Yes
`host_instructions`	No	Set only when backend_used is 'host': the improved_prompt is a template draft, and these are the steps the calling model should take to finish it.
`clarifying_questions`	Yes

Schema Changelog

Changes observed during successful MCP inspections.

First observedv0.1.0Oct 15, 2026

TDQS

A4.1/5.0

Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations exist, so the description carries the full burden, and it discloses meaningful behavior: the statelessness constraint, the four-part return payload, and the degraded 'host' path where improved_prompt is only a template requiring host_instructions. It omits cost/latency, credential requirements, and determinism concerns, so it is strong but not exhaustive.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loads the purpose, then moves through usage constraint, return shape, and edge case in four dense sentences. The long list of nine-plus technique names is the one slightly bloated element, though it justifies the tool's scope.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

With an output schema present, the return explanation is a bonus rather than a requirement, and the description goes further by covering the host-backend edge case the schema only implies. For an 8-parameter tool with one required field, nothing needed to call it correctly is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the baseline is 3. The description still adds value by tying prompt/context/history together into one verbatim-passing rule and by naming 'backend_used' and the host-draft fallback, giving the backend parameter practical meaning beyond the enum values.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The opening sentence gives a concrete verb and object ('Rewrite a prompt so a model answers it better') and names the concrete technique families applied, so the agent knows exactly what the tool produces. It never differentiates itself from the sibling 'analyze_prompt', so the boundary is left to the name alone.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives clear operational guidance for invocation: 'The server cannot see the conversation, so pass the current prompt verbatim plus the earlier turns and context it depends on.' That is a real usage condition an agent must know. It stops short of naming analyze_prompt/list_techniques as alternatives or saying when not to call it.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server