Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@PromptSmith Recommend a reasoning strategy for architecting a global supply chain"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Reasoning as a Service
An MCP server that analyzes any task, selects the optimal reasoning framework from 47 peer-reviewed strategies, and generates a tailored meta-prompt -- ready to feed to any LLM. No LLM calls required for selection. Deterministic. Sub-millisecond.
Most AI agents use Chain of Thought for everything. That's like using a hammer for every job.
A code generation task needs a different reasoning strategy than a research synthesis task, which needs a different strategy than a logic puzzle. The academic literature describes over 45 distinct reasoning frameworks -- each optimized for specific task types and complexity levels. But no developer has time to read dozens of papers and manually select the right one for every prompt.
PromptCore encodes that expertise into a single tool call.
Related MCP server: Athena MCP
flowchart LR
A["Task Input"] --> B["Category Detection"]
B --> C["Complexity Scoring"]
C --> D["Intent Analysis"]
D --> E["Framework Selection"]
E --> F["Meta-Prompt Generation"]
style A fill:#1a1a2e,stroke:#e94560,color:#fff
style B fill:#1a1a2e,stroke:#0f3460,color:#fff
style C fill:#1a1a2e,stroke:#0f3460,color:#fff
style D fill:#1a1a2e,stroke:#0f3460,color:#fff
style E fill:#1a1a2e,stroke:#e94560,color:#fff
style F fill:#1a1a2e,stroke:#16c79a,color:#fffThree steps, one tool call:
Send your task -- PromptCore analyzes category (code, math, logic, creative, research, data, planning), complexity (0--10), and intent (17 types including decomposition, verification, exploration)
Framework selected -- Heuristic scoring matches your task against 47 peer-reviewed reasoning strategies, selecting the one with the highest fit
Meta-prompt returned -- A structured prompt, built on the selected framework's methodology, ready to feed to any LLM
No LLM calls. No API keys for selection. Runs in under 1ms.
from promptcore.domain import FrameworkSelector, PromptBuilder
selector = FrameworkSelector()
analysis = selector.analyze("Write a recursive function to calculate fibonacci numbers")
print(analysis.category) # CODE
print(analysis.complexity_score) # 3.3
print(analysis.recommended_framework) # program_of_thoughtsTask: "Write a recursive function to calculate fibonacci numbers"
Category: CODE
Complexity: 3.3 / 10
Framework: Program of Thoughts
Meta-prompt:
"Approach this problem by expressing your reasoning as executable code.
Break the task into computational steps:
1. Define the problem as a function signature
2. Identify the base cases and recursive structure
3. Express the logic as working code with comments explaining each decision
4. Verify correctness by tracing through example inputs
..."Compare that to a naive prompt ("Write a fibonacci function") or a blanket Chain of Thought ("Think step by step..."). PromptCore selects Program of Thoughts because the task is code-category with moderate complexity -- a framework specifically designed for tasks where reasoning is best expressed as executable logic.
PromptCore includes 47 reasoning frameworks from published research, organized into seven categories.
graph TB
subgraph ZS["Zero-Shot"]
zs1["Role Prompting"]
zs2["Emotion Prompting"]
zs3["System-2 Attention"]
zs4["SimToM"]
zs5["Rephrase & Respond"]
zs6["Self-Ask"]
end
subgraph TG["Thought Generation"]
tg1["Chain of Thought"]
tg2["Step-Back"]
tg3["Thread of Thought"]
tg4["Tab-CoT"]
tg5["Contrastive CoT"]
tg6["Complexity-Based"]
tg7["Active Prompting"]
tg8["Analogical"]
tg9["Directional Stimulus"]
end
subgraph DC["Decomposition"]
dc1["Tree of Thoughts"]
dc2["Least-to-Most"]
dc3["Program of Thoughts"]
dc4["Skeleton of Thought"]
dc5["Plan-and-Solve"]
dc6["Faithful CoT"]
dc7["Recursion of Thought"]
end
subgraph EN["Ensembling"]
en1["Self-Consistency"]
en2["DENSE"]
en3["MoRE"]
en4["Meta-CoT"]
en5["Prompt Paraphrasing"]
end
subgraph SC["Self-Criticism"]
sc1["Reflexion"]
sc2["Maieutic"]
sc3["Chain of Verification"]
sc4["Self-Refine"]
sc5["Self-Calibration"]
sc6["Reverse CoT"]
sc7["Cumulative Reasoning"]
end
subgraph AD["Advanced"]
ad1["ReAct"]
ad2["Graph of Thoughts"]
ad3["Reasoning via Planning"]
ad4["Chain of Density"]
ad5["Buffer of Thoughts"]
ad6["Chain of Table"]
end
subgraph MD["Modern"]
md1["Self-Discover"]
md2["Chain of Draft"]
md3["CRITIC"]
md4["Chain of Code"]
md5["RE2 Re-Reading"]
md6["Chain-of-Abstraction"]
md7["Deliberate-then-Generate"]
end
style ZS fill:#0d1117,stroke:#58a6ff,color:#c9d1d9
style TG fill:#0d1117,stroke:#f0883e,color:#c9d1d9
style DC fill:#0d1117,stroke:#a371f7,color:#c9d1d9
style EN fill:#0d1117,stroke:#3fb950,color:#c9d1d9
style SC fill:#0d1117,stroke:#f85149,color:#c9d1d9
style AD fill:#0d1117,stroke:#d2a8ff,color:#c9d1d9
style MD fill:#0d1117,stroke:#e3b341,color:#c9d1d9Framework | Best For | Complexity Threshold |
Role Prompting | Creative, General, Research | 1.0 |
Emotion Prompting | Creative, General | 1.0 |
Rephrase and Respond | General, Research | 2.0 |
Chain of Thought | Math, Logic, Code | 2.0 |
System 2 Attention | Logic, Research, General | 3.0 |
Thread of Thought | Research, Data, General | 3.0 |
Tab-CoT | Data, Math, Logic | 3.0 |
Directional Stimulus | Creative, General | 3.0 |
Skeleton of Thought | Creative, General, Planning | 3.0 |
Self-Calibration | Math, Logic, General | 3.0 |
Chain of Density | Research, Data, General | 3.0 |
Prompt Paraphrasing | General, Logic | 3.0 |
Sim-to-M | Logic, General | 4.0 |
Self-Ask | Research, Logic, General | 4.0 |
Step Back | Research, Logic, General | 4.0 |
Analogical | General, Creative, Code | 4.0 |
Program of Thoughts | Math, Code, Data | 4.0 |
Plan and Solve | Planning, Code, Math | 4.0 |
Self-Consistency | Math, Logic | 4.0 |
Self-Refine | Creative, Code, General | 4.0 |
Chain of Table | Data | 4.0 |
Least to Most | Code, Math, Planning | 5.0 |
Contrastive CoT | Math, Logic, Code | 5.0 |
Active Prompting | General, Research, Logic | 5.0 |
Faithful CoT | Math, Logic, Code | 5.0 |
Demonstration Ensembling | General, Data, Logic | 5.0 |
Maieutic | Research, Logic, General | 5.0 |
Chain of Verification | Research, General, Data | 5.0 |
Reverse CoT | Math, Logic, Code | 5.0 |
Buffer of Thoughts | General, Math, Code | 5.0 |
Complexity-Based | Math, Logic | 6.0 |
Tree of Thoughts | Creative, Planning, Research | 6.0 |
Mixture of Reasoning | General, Research, Logic | 6.0 |
Meta-CoT | Logic, Math, Research | 6.0 |
Cumulative Reasoning | Logic, Math, Research | 6.0 |
Recursion of Thought | Math, Code, Logic | 7.0 |
Self-Discover | General, Logic, Planning, Research | 7.0 |
CRITIC | Research, Data, Code, General | 6.0 |
Chain of Code | Code, Math, Data, Logic | 6.0 |
Chain-of-Abstraction | Math, Data, Research | 5.0 |
Deliberate-then-Generate | Creative, General, Research | 4.0 |
RE2 (Re-Reading) | Math, Logic, Research | 2.0 |
Chain of Draft | Math, Logic, Code, Creative, General | 3.0 |
ReAct | Research, Code, Data | 7.0 |
Reflexion | Code, Math, Logic | 8.0 |
Graph of Thoughts | Planning, Research, Logic | 8.0 |
Reasoning via Planning | Planning, Logic, Code | 8.0 |
Tool | Description |
| Analyze a task and recommend the optimal reasoning framework with category, complexity, and intent breakdown |
| Generate a structured meta-prompt using the selected framework |
| Record feedback about prompt effectiveness for analytics |
| Enumerate all 47 frameworks with metadata |
| Query usage statistics and framework effectiveness trends |
graph TB
MCP["MCP Interface<br/><code>main.py</code>"] --> Domain
subgraph Domain["Domain Layer"]
Selector["FrameworkSelector<br/><code>selector.py</code>"]
Frameworks["Framework Registry<br/><code>frameworks/</code> package<br/>47 implementations"]
Builder["PromptBuilder<br/><code>builder.py</code>"]
Selector --> Frameworks
Frameworks --> Builder
end
subgraph Persistence["Persistence Layer"]
Models["SQLAlchemy Models<br/><code>models.py</code>"]
Storage["SQLite Storage<br/><code>storage.py</code>"]
Models --> Storage
end
subgraph Utils["Utilities"]
Complexity["Complexity Analyzer<br/><code>complexity.py</code>"]
end
Domain --> Persistence
Selector --> Complexity
style MCP fill:#1a1a2e,stroke:#e94560,color:#fff
style Domain fill:#0d1117,stroke:#58a6ff,color:#c9d1d9
style Persistence fill:#0d1117,stroke:#3fb950,color:#c9d1d9
style Utils fill:#0d1117,stroke:#f0883e,color:#c9d1d9src/promptcore/
├── main.py # MCP server entry point (FastMCP, stdio transport)
├── domain/
│ ├── frameworks/ # 47 reasoning framework implementations
│ ├── selector.py # Task analysis: category, complexity, intent, framework scoring
│ └── builder.py # Meta-prompt assembly from framework templates
├── persistence/
│ ├── models.py # SQLAlchemy + Pydantic models (ReasoningLog)
│ └── storage.py # SQLite operations for traces and analytics
└── utils/
└── complexity.py # Text complexity analysis (token count, keywords, ambiguity)PromptCore's category detection layer ensures that adversarial or ambiguous prompts are routed to appropriate reasoning frameworks. Tasks with mixed signals (e.g., a prompt that looks like code but contains social engineering patterns) are scored conservatively -- defaulting to frameworks with built-in verification steps (Chain of Verification, Self-Calibration) rather than naive execution frameworks. This makes PromptCore a safer default for agent pipelines processing untrusted user input.
# 1. Clone
git clone https://github.com/BlinkVoid/PromptCore.git
cd PromptCore
# 2. Install dependencies
uv sync
# 3. Run the MCP server
uv run python -m promptcore.mainRequirements: Python 3.10+, uv
Add PromptCore to your MCP client. Works with Claude Code, Cline, Continue, Cursor, and any MCP-compatible agent.
.mcp.json in project root){
"mcpServers": {
"promptcore": {
"command": "uv",
"args": ["run", "--directory", "/path/to/PromptCore", "python", "-m", "promptcore.main"],
"env": {
"PYTHONPATH": "/path/to/PromptCore/src",
"UV_LINK_MODE": "copy",
"FASTMCP_SHOW_STARTUP_BANNER": "false"
}
}
}
}claude_desktop_config.json){
"mcpServers": {
"promptcore": {
"command": "uv",
"args": ["run", "--directory", "/path/to/PromptCore", "python", "-m", "promptcore.main"],
"env": {
"PYTHONPATH": "/path/to/PromptCore/src"
}
}
}
}# Any MCP client can call PromptCore as a tool
result = mcp_call("promptcore", "generate_meta_prompt", {
"task": "Analyze the trade-offs between microservices and monolith for a fintech startup",
"context": "The team has 4 engineers and needs to ship in 3 months"
})
# Feed the optimized prompt to any LLM
response = llm.generate(result["meta_prompt"])from promptcore.domain import FrameworkSelector, PromptBuilder
selector = FrameworkSelector()
builder = PromptBuilder()
# Analyze
analysis = selector.analyze("Design a distributed cache invalidation strategy")
print(f"Category: {analysis.category}") # PLANNING
print(f"Complexity: {analysis.complexity_score}") # ~7.2
print(f"Framework: {analysis.recommended_framework}") # reasoning_via_planning
# Generate
result = builder.build(analysis.task, analysis=analysis)
print(result.meta_prompt) # Structured prompt using Reasoning-via-Planning methodology# Middleware pattern: enhance every LLM call with optimal reasoning
def enhanced_llm_call(task: str, context: str = "") -> str:
# Step 1: Get optimal reasoning strategy
strategy = mcp_call("promptcore", "recommend_strategy", {"task": task})
# Step 2: Generate meta-prompt
prompt = mcp_call("promptcore", "generate_meta_prompt", {
"task": task,
"context": context,
"framework": strategy["recommended_framework"]
})
# Step 3: Execute with any LLM
result = llm.generate(prompt["meta_prompt"])
# Step 4: Log feedback for analytics
mcp_call("promptcore", "log_execution_feedback", {
"log_id": prompt["log_id"],
"feedback": "success",
"notes": "Output matched expected format"
})
return resultPromptCore is not the only approach to prompt optimization. See docs/COMPARISON.md for detailed comparisons against DSPy (Stanford), PromptFlow (Microsoft), LangChain Templates, manual engineering, and interactive playgrounds -- including an honest assessment of where each approach wins.
The short version: PromptCore is the only tool that provides automatic, zero-LLM-cost framework selection from a curated library of 47 peer-reviewed strategies, exposed as MCP tools. If you need LLM-in-the-loop optimization, use DSPy. If you need a visual workflow, use PromptFlow. If you want drop-in reasoning enhancement for agent pipelines, use PromptCore.
Contributions are welcome. See CONTRIBUTING.md for setup instructions and guidelines.
Areas where contributions would have the most impact:
New reasoning framework implementations (with paper citations)
Improved complexity scoring heuristics
Benchmark results validating framework selection quality
Integrations with additional MCP clients
Generate an optimized meta-prompt for the given task.
Analyzes the task, selects the best reasoning framework (or uses the specified one), and generates a structured prompt designed to elicit high-quality reasoning.
| Name | Required | Description | Default |
|---|---|---|---|
| task | Yes | The task or question to create a prompt for | |
| context | No | Additional context to include in the prompt | |
| framework | No | Specific framework to use (optional, auto-selects if not provided) | |
| persist | No | Whether to log this generation for analytics |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It mentions analyzing tasks and generating structured prompts, but lacks details on permissions, rate limits, side effects, or what 'optimized' entails. For a tool with no annotation coverage, this leaves significant gaps in understanding its behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise and well-structured in two sentences. The first sentence states the core purpose, and the second elaborates on the process. There's no wasted text, though it could be slightly more front-loaded with key details.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (generating prompts with frameworks) and the presence of an output schema, the description is moderately complete. It covers the purpose but lacks behavioral details and usage guidelines. With no annotations and incomplete contextual guidance, it's adequate but has clear gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents all parameters thoroughly. The description adds minimal value beyond the schema, mentioning 'task' and 'framework' implicitly but not explaining their semantics further. With high schema coverage, the baseline score of 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Generate an optimized meta-prompt for the given task.' It specifies the verb ('generate') and resource ('optimized meta-prompt'), and mentions analyzing tasks and selecting reasoning frameworks. However, it doesn't explicitly differentiate from sibling tools like 'recommend_strategy' which might have overlapping functionality.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. It doesn't mention sibling tools like 'list_available_frameworks' or 'recommend_strategy', nor does it specify prerequisites or contexts where this tool is preferred. Usage is implied but not explicitly stated.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Get usage statistics for PromptCore.
Shows total prompts generated, distribution by framework and category, and average complexity scores.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It states what data is returned but doesn't cover important aspects like whether this is a read-only operation, if it requires authentication, rate limits, freshness of data, or error conditions. For a statistics tool with zero annotation coverage, this leaves significant gaps.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is appropriately concise with three sentences that each add value: stating the tool's purpose, listing key statistics categories, and specifying metrics. It's front-loaded with the core purpose and avoids unnecessary elaboration. Minor deduction for slightly repetitive phrasing ('distribution by framework and category' could be more streamlined).
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given that the tool has no parameters, has an output schema (so return values are documented elsewhere), and provides basic statistics, the description is minimally complete. However, without annotations and with sibling tools that might overlap (like log_execution_feedback), more context about when this tool is appropriate would improve completeness for the agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has 0 parameters with 100% schema description coverage, so the schema already fully documents the input requirements. The description appropriately doesn't repeat parameter information, maintaining a baseline of 4 for parameterless tools that don't waste space on nonexistent parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose with specific verbs ('Get usage statistics') and resources ('for PromptCore'), and lists the types of statistics provided (total prompts, distribution by framework/category, average complexity). It doesn't explicitly differentiate from sibling tools, but since no siblings appear to provide similar statistics, this is adequate.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. It doesn't mention prerequisites, frequency of use, or how it relates to sibling tools like log_execution_feedback or recommend_strategy. The agent must infer usage context solely from the purpose.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List all available reasoning frameworks with their descriptions.
Use this to understand what frameworks are available and when each is best used.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It describes the tool as a list operation, implying it's read-only and non-destructive, which is helpful. However, it lacks details on behavioral traits like rate limits, authentication needs, pagination, or response format. The description adds basic context but doesn't fully compensate for the absence of annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, front-loaded with the core purpose and followed by usage guidance. Every sentence adds value without redundancy, making it efficient and well-structured. There's no wasted text, and it's appropriately sized for the tool's complexity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has 0 parameters, no annotations, and an output schema exists (which handles return values), the description is reasonably complete. It covers the purpose and usage context adequately. However, it could be more comprehensive by addressing potential behavioral aspects like error handling or data freshness, but the output schema reduces the need for extensive detail.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 0 parameters with 100% coverage, so no parameter documentation is needed. The description doesn't add parameter details, which is appropriate. A baseline of 4 is applied for tools with no parameters, as there's nothing to compensate for, and the description doesn't introduce confusion.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose with a specific verb ('List') and resource ('all available reasoning frameworks with their descriptions'). It distinguishes itself from siblings like 'generate_meta_prompt' or 'recommend_strategy' by focusing on enumeration rather than generation or recommendation. However, it doesn't explicitly contrast with 'get_usage_stats' or 'log_execution_feedback', which might also involve listing data.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides clear usage context: 'Use this to understand what frameworks are available and when each is best used.' This gives a specific scenario for when to use the tool. However, it doesn't explicitly state when NOT to use it or name alternatives among the sibling tools, such as using 'recommend_strategy' for selection instead of just listing.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Log feedback about how a generated prompt performed.
Used to track effectiveness and improve framework selection over time.
| Name | Required | Description | Default |
|---|---|---|---|
| task_id | Yes | The task ID from generate_meta_prompt | |
| feedback | Yes | Feedback about how well the prompt worked |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It mentions the tool logs feedback for tracking and improvement, but doesn't disclose behavioral traits like whether it's idempotent, requires specific permissions, has rate limits, or what happens on failure. For a logging tool with zero annotation coverage, this is a significant gap in transparency.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is appropriately sized with two sentences. The first sentence states the purpose, and the second provides usage context. There's no wasted text, but it could be slightly more front-loaded with key details like the connection to generate_meta_prompt.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's moderate complexity (logging with two parameters), 100% schema coverage, and the presence of an output schema (which handles return values), the description is reasonably complete. It covers purpose and high-level usage, but lacks behavioral details that annotations would normally provide, such as idempotency or error handling.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents both parameters (task_id and feedback) with descriptions. The description adds no additional meaning beyond what the schema provides, such as format examples or constraints. Baseline 3 is appropriate when the schema does the heavy lifting.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Log feedback about how a generated prompt performed' specifies the verb (log) and resource (feedback). It distinguishes from siblings like generate_meta_prompt (creates prompts) and get_usage_stats (retrieves statistics). However, it doesn't explicitly differentiate from all siblings (e.g., recommend_strategy could involve feedback).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage context: 'Used to track effectiveness and improve framework selection over time' suggests it's for post-generation evaluation. It references generate_meta_prompt via the task_id parameter, but doesn't explicitly state when to use this tool versus alternatives like get_usage_stats or recommend_strategy for tracking purposes.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Analyze a task and recommend the optimal reasoning framework.
Returns the detected category, complexity_score, recommended framework, and alternative options.
| Name | Required | Description | Default |
|---|---|---|---|
| task | Yes | The task or question to analyze | |
| context | No | Additional context about the task |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It states the tool returns a category, complexity_score, recommended framework, and alternatives, but doesn't cover critical aspects like whether this is a read-only analysis (implied but not stated), computational cost, rate limits, or error conditions. This leaves significant gaps for a tool that performs analysis.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise and front-loaded, with the core purpose stated first. The second sentence efficiently lists return values. However, it could be slightly more structured by explicitly separating purpose from output, and it includes a minor redundancy ('optimal' and 'recommended' overlap), preventing a perfect score.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's moderate complexity (analysis with two parameters), no annotations, and the presence of an output schema (which handles return value documentation), the description is reasonably complete. It covers the purpose and output at a high level, but lacks behavioral details like error handling or performance characteristics, which holds it back from a score of 5.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema description coverage is 100%, so the schema fully documents the two parameters ('task' and 'context'). The description adds no additional meaning beyond what's in the schema (e.g., it doesn't explain what constitutes a 'task' or how 'context' influences the analysis), resulting in the baseline score of 3.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Analyze a task and recommend the optimal reasoning framework.' It specifies the verb ('analyze' and 'recommend') and resource ('task' and 'reasoning framework'), but doesn't explicitly differentiate it from sibling tools like 'list_available_frameworks' or 'generate_meta_prompt', which prevents a score of 5.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. It doesn't mention sibling tools like 'list_available_frameworks' (which might list frameworks without analysis) or 'generate_meta_prompt' (which might use a framework), leaving the agent with no context for tool selection.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0generate_meta_promptget_usage_statslist_available_frameworkslog_execution_feedbackrecommend_strategyScored across 5 tools
Each tool has a clearly distinct purpose with no overlap. generate_meta_prompt creates prompts, get_usage_stats retrieves statistics, list_available_frameworks enumerates options, log_execution_feedback captures performance data, and recommend_strategy provides analysis. An agent can easily distinguish between these five distinct functions.
All tools follow a consistent verb_noun pattern with snake_case throughout. The naming is predictable and readable: generate_meta_prompt, get_usage_stats, list_available_frameworks, log_execution_feedback, and recommend_strategy. There are no deviations in style or convention.
With 5 tools, this server is well-scoped for its purpose of prompt generation and management. Each tool earns its place by covering distinct aspects: generation, statistics, listing, feedback, and strategy recommendation. This count is neither too thin nor too heavy for the domain.
The tool surface covers the core lifecycle of prompt generation and optimization well, including creation (generate_meta_prompt), analysis (recommend_strategy), monitoring (get_usage_stats, log_execution_feedback), and discovery (list_available_frameworks). A minor gap is the lack of tools for editing or deleting prompts, but agents can work around this by regenerating prompts as needed.
MCP server for building and testing AI agents with multi-model experimentation and insights.
MCP server connecting AI agents to 100+ apps (Gmail, Slack, Notion, GitHub) via one-click OAuth.
MCP server unifying ERPs, CRMs, APIs and knowledge base for Claude, ChatGPT and Gemini.
MCP Server for an Agent Task Marketplace
