MCP_Atom_of_Thoughts

MCP_Atom_of_Thoughts

3.5

If you are the rightful owner of MCP_Atom_of_Thoughts and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

A Model Context Protocol (MCP) server implementation of Atom of Thoughts, a decomposition-based reasoning framework.

Atom of Thoughts (AoT)

A Model Context Protocol (MCP) server implementation of Atom of Thoughts, a decomposition-based reasoning framework.

Note: This implementation is based on the research paper "Atom of Thoughts for Markov LLM Test-Time Scaling" (Teng et al., 2025).

MCP.so

ํ•œ๊ตญ์–ด ์„ค๋ช…

English Documentation

This repository implements Atom of Thoughts (AoT), a decomposition-based reasoning framework, as a Model Context Protocol (MCP) server. This implementation is based on the concepts presented in the paper "Atom of Thoughts for Markov LLM Test-Time Scaling" (Teng et al., 2025).

Available Tools

Two main tools are provided:

  1. AoT (Full Version): A complete Atom of Thoughts tool with full capabilities for deep analysis and complex problem solving.
  2. AoT-light (Lightweight Version): A streamlined version optimized for faster processing and quicker results.

AoT-light: Lightweight Version

AoT-light is designed for faster processing in time-sensitive situations:

  • Key Features:

    • Lower maximum depth (3 instead of 5) for faster processing
    • Simplified verification process
    • Immediate conclusion suggestion for high-confidence hypotheses
    • Reduced computational overhead and response payload
    • Optimized for speed rather than exhaustive analysis
  • Use Cases:

    • Quick brainstorming sessions requiring atomic thought organization
    • Time-sensitive problem solving where speed is prioritized over exhaustive analysis
    • Simpler reasoning tasks that don't require deep decomposition
    • Initial exploration before using the full AoT for deeper analysis
    • Learning or demonstration purposes where response time is important

Use Cases

Atom of Thoughts is effective in the following scenarios:

  • Solving problems requiring complex reasoning
  • Generating hypotheses that need verification from multiple perspectives
  • Deriving high-confidence conclusions in scenarios where accuracy is crucial
  • Minimizing logical errors in critical tasks
  • Decision-making requiring multiple verification steps

Atom Types

AoT uses five types of atoms:

  1. premise: Basic assumptions or given information for problem solving
  2. reasoning: Logical reasoning process based on other atoms
  3. hypothesis: Proposed solutions or intermediate conclusions
  4. verification: Process to evaluate the validity of other atoms (especially hypotheses)
  5. conclusion: Verified hypotheses or final problem solutions

Core Features

1. Decomposition-Contraction Mechanism

A mechanism to decompose atoms into smaller sub-atoms and contract them back after verification.

  • Decomposition: Breaking complex atoms into smaller sub-atoms.

    • startDecomposition(atomId): Start atom decomposition
    • addToDecomposition(decompositionId, atomId): Add sub-atom to decomposition
    • completeDecomposition(decompositionId): Complete decomposition process
  • Contraction: Contract back to the original atom once all sub-atoms are verified.

    • Calculate confidence of the original atom based on sub-atoms' confidence levels
    • Automatically suggest conclusions for high-confidence verified hypotheses
2. Automatic Termination Mechanism
  • Automatically terminates when reaching maximum depth or finding a high-confidence conclusion.
  • getTerminationStatus(): Return current termination status and reason
  • getBestConclusion(): Return the conclusion with highest confidence

Parameter Descriptions

  • atomId: Unique identifier for the atom (e.g., 'A1', 'H2')
  • content: Actual content of the atom
  • atomType: Type of atom (one of: premise, reasoning, hypothesis, verification, conclusion)
  • dependencies: List of IDs of other atoms this atom depends on
  • confidence: Confidence level of this atom (value between 0-1)
  • isVerified: Whether this atom has been verified
  • depth: Depth level of this atom in the decomposition-contraction process

Usage Method

  1. Understand the problem and define necessary premise atoms
  2. Create reasoning atoms based on premises
  3. Create hypothesis atoms based on reasoning
  4. Create verification atoms to verify hypotheses
  5. Derive conclusion atoms based on verified hypotheses
  6. Use atom decomposition to explore deeper when necessary
  7. Present the high-confidence conclusion atom as the final answer

Comparing Sequential Thinking and Atom of Thoughts (More Testing Needed)

After applying both thinking tools to the same topic, the following differences and performance characteristics were observed:

Structural Differences

Sequential Thinking:

  • Linear thinking process: progresses sequentially from one thought to the next
  • Predicts the total number of thoughts in advance
  • Each thinking stage is built upon previous stages

Atom of Thoughts:

  • Non-linear, network structure: multiple thought units (atoms) interconnect with dependencies
  • Forms systematic structure according to atom types (premise, reasoning, hypothesis, verification, conclusion)
  • Explicitly evaluates the confidence level of each atom
Comparative Strengths

Sequential Thinking Strengths:

  • Intuitive flow: similar to natural human thinking processes
  • Simplicity: simple structure allows quick application to straightforward problems
  • Flexibility: can modify previous stages or change direction during the thinking process

Atom of Thoughts Strengths:

  • Confidence evaluation: explicitly measures the confidence of each thought to improve conclusion validity
  • Verification process: evaluates hypotheses through systematic verification stages
  • Dependency tracking: clearly tracks which premises or reasoning influenced specific conclusions
  • Parallel processing: can consider multiple thought atoms simultaneously
Efficiency and Accuracy

Efficiency:

  • Sequential Thinking: more efficient for simple problems, with faster progression of thought
  • Atom of Thoughts: more efficient for complex problems, but has initial overhead in building systematic structures

Accuracy:

  • Sequential Thinking: possibility of error accumulation from previous stages as the thinking process deepens
  • Atom of Thoughts: reduced error possibility through verification stages and confidence assessment, leading to more reliable conclusions
Suitability by Purpose

Cases Suitable for Sequential Thinking:

  • Simple to moderately complex problems
  • Time-constrained situations
  • When natural storytelling or explanation is needed

Cases Suitable for Atom of Thoughts:

  • Highly complex problems
  • Situations where accuracy and reliability are crucial
  • Hypotheses requiring verification from multiple perspectives
  • Reasoning with complex dependency relationships
Conclusion

Both tools can contribute to improving artificial intelligence's reasoning abilities, but the appropriate tool varies depending on the nature of the problem and requirements. Sequential Thinking is useful when intuitive and quick thinking processes are needed, while Atom of Thoughts is more suitable for complex problems requiring systematic verification and high reliability.

Command Tool (atomcommands)

A command tool to control the decomposition-contraction mechanism and automatic termination of Atom of Thoughts.

Available Commands:

  1. decompose: Decompose a specified atom into smaller sub-atoms
    • Required parameter: atomId
  2. complete_decomposition: Complete an ongoing decomposition process
    • Required parameter: decompositionId
  3. termination_status: Check the termination status of the current AoT process
  4. best_conclusion: Get the verified conclusion with the highest confidence
  5. set_max_depth: Change the maximum depth limit
    • Required parameter: maxDepth

Installing via Smithery

To install Atom of Thoughts for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @kbsooo/mcp_atom_of_thoughts --client claude

MCP Server Configuration

To use the Atom of Thoughts MCP server, you need to register it in your Claude Desktop or Cline MCP settings. Here is an example configuration:

{ 
  "mcpServers": { 
    "atom-of-thoughts": { 
      "command": "node", 
      "args": ["/ABSOLUTE/PATH/TO/PARENT/FOLDER/atom-of-thoughts/build/index.js"], 
      "disabled": false, 
      "autoApprove": [] 
    } 
  } 
}

Replace /ABSOLUTE/PATH/TO/PARENT/FOLDER with the actual absolute path to the project on your system. After saving the configuration, restart Claude Desktop or Cline to use the Atom of Thoughts MCP server.

For detailed implementation and code-level documentation, please refer to the source code in this repository.

ํ•œ๊ตญ์–ด ์„ค๋ช…

Atom of Thoughts๋ž€?

Atom of Thoughts(AoT)๋Š” ๋ณต์žกํ•œ ๋ฌธ์ œ๋ฅผ ๋…๋ฆฝ์ ์ด๊ณ  ์žฌ์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ์›์ž ๋‹จ์œ„์˜ ์‚ฌ๊ณ ๋กœ ๋ถ„ํ•ดํ•˜์—ฌ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๋Š” ๋„๊ตฌ์ž…๋‹ˆ๋‹ค. ์ด ๋„๊ตฌ๋Š” ์ „ํ†ต์ ์ธ ์ˆœ์ฐจ์  ์‚ฌ๊ณ  ๋ฐฉ์‹๊ณผ ๋‹ฌ๋ฆฌ, ์‚ฌ๊ณ ์˜ ๊ธฐ๋ณธ ๋‹จ์œ„์ธ '์›์ž'๋“ค์ด ์„œ๋กœ ์˜์กด์„ฑ์„ ๊ฐ–๊ณ  ๊ตฌ์„ฑ๋˜์–ด ๋” ๊ฐ•๋ ฅํ•œ ๋ฌธ์ œ ํ•ด๊ฒฐ์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•ฉ๋‹ˆ๋‹ค. ์ด ๊ตฌํ˜„์€ "Atom of Thoughts for Markov LLM Test-Time Scaling"(Teng et al., 2025) ๋…ผ๋ฌธ์˜ ๊ฐœ๋…์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค.

์ œ๊ณต๋˜๋Š” ๋„๊ตฌ

ํ˜„์žฌ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋‘ ๊ฐ€์ง€ ์ฃผ์š” ๋„๊ตฌ๊ฐ€ ์ œ๊ณต๋ฉ๋‹ˆ๋‹ค:

  1. AoT (์ „์ฒด ๋ฒ„์ „): ์‹ฌ์ธต์ ์ธ ๋ถ„์„๊ณผ ๋ณต์žกํ•œ ๋ฌธ์ œ ํ•ด๊ฒฐ์„ ์œ„ํ•œ ์™„์ „ํ•œ ๊ธฐ๋Šฅ์„ ๊ฐ–์ถ˜ Atom of Thoughts ๋„๊ตฌ์ž…๋‹ˆ๋‹ค.
  2. AoT-light (๊ฒฝ๋Ÿ‰ ๋ฒ„์ „): ๋” ๋น ๋ฅธ ์ฒ˜๋ฆฌ์™€ ์‹ ์†ํ•œ ๊ฒฐ๊ณผ๋ฅผ ์œ„ํ•ด ์ตœ์ ํ™”๋œ ๊ฒฝ๋Ÿ‰ ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.

AoT-light: ๊ฒฝ๋Ÿ‰ ๋ฒ„์ „

AoT-light๋Š” ์‹œ๊ฐ„์ด ์ค‘์š”ํ•œ ์ƒํ™ฉ์—์„œ ๋” ๋น ๋ฅธ ์ฒ˜๋ฆฌ๋ฅผ ์œ„ํ•ด ์„ค๊ณ„๋œ ๊ฒฝ๋Ÿ‰ ๋ฒ„์ „์ž…๋‹ˆ๋‹ค:

  • ์ฃผ์š” ํŠน์ง•:

    • ๋‚ฎ์€ ์ตœ๋Œ€ ๊นŠ์ด (5 ๋Œ€์‹  3) ์„ค์ •์œผ๋กœ ๋น ๋ฅธ ์ฒ˜๋ฆฌ
    • ๊ฐ„์†Œํ™”๋œ ๊ฒ€์ฆ ํ”„๋กœ์„ธ์Šค
    • ๋†’์€ ์‹ ๋ขฐ๋„์˜ ๊ฐ€์„ค์— ๋Œ€ํ•œ ์ฆ‰๊ฐ์ ์ธ ๊ฒฐ๋ก  ์ œ์•ˆ
    • ์ถ•์†Œ๋œ ๊ณ„์‚ฐ ์˜ค๋ฒ„ํ—ค๋“œ ๋ฐ ์‘๋‹ต ๋ฐ์ดํ„ฐ
    • ์ฒ ์ €ํ•œ ๋ถ„์„๋ณด๋‹ค ์†๋„์— ์ตœ์ ํ™”
  • ์‚ฌ์šฉ ์‹œ๋‚˜๋ฆฌ์˜ค:

    • ์›์ž์  ์‚ฌ๊ณ  ๊ตฌ์„ฑ์ด ํ•„์š”ํ•œ ๋น ๋ฅธ ๋ธŒ๋ ˆ์ธ์Šคํ† ๋ฐ ์„ธ์…˜
    • ์ฒ ์ €ํ•œ ๋ถ„์„๋ณด๋‹ค ์†๋„๊ฐ€ ์ค‘์š”ํ•œ ์‹œ๊ฐ„์— ๋ฏผ๊ฐํ•œ ๋ฌธ์ œ ํ•ด๊ฒฐ
    • ๊นŠ์€ ๋ถ„ํ•ด๊ฐ€ ํ•„์š”ํ•˜์ง€ ์•Š์€ ๋‹จ์ˆœํ•œ ์ถ”๋ก  ์ž‘์—…
    • ์ „์ฒด AoT๋ฅผ ์‚ฌ์šฉํ•œ ์‹ฌ์ธต ๋ถ„์„ ์ „ ์ดˆ๊ธฐ ํƒ์ƒ‰
    • ์‘๋‹ต ์‹œ๊ฐ„์ด ์ค‘์š”ํ•œ ํ•™์Šต ๋˜๋Š” ์‹œ์—ฐ ๋ชฉ์ 

์‚ฌ์šฉ ์‹œ๋‚˜๋ฆฌ์˜ค

๋‹ค์Œ๊ณผ ๊ฐ™์€ ๊ฒฝ์šฐ์— Atom of Thoughts๋ฅผ ์‚ฌ์šฉํ•˜๋ฉด ํšจ๊ณผ์ ์ž…๋‹ˆ๋‹ค:

  • ๋ณต์žกํ•œ ์ถ”๋ก ์ด ํ•„์š”ํ•œ ๋ฌธ์ œ ํ•ด๊ฒฐ
  • ์—ฌ๋Ÿฌ ๊ด€์ ์—์„œ ๊ฒ€์ฆ์ด ํ•„์š”ํ•œ ๊ฐ€์„ค ์ƒ์„ฑ
  • ์ •ํ™•๋„๊ฐ€ ์ค‘์š”ํ•œ ๋ฌธ์ œ์—์„œ ์‹ ๋ขฐ๋„ ๋†’์€ ๊ฒฐ๋ก  ๋„์ถœ
  • ๋…ผ๋ฆฌ์  ์˜ค๋ฅ˜๋ฅผ ์ตœ์†Œํ™”ํ•ด์•ผ ํ•˜๋Š” ์ž‘์—…
  • ์—ฌ๋Ÿฌ ๋‹จ๊ณ„์˜ ๊ฒ€์ฆ์ด ํ•„์š”ํ•œ ์˜์‚ฌ๊ฒฐ์ •

์›์ž ์œ ํ˜•

Atom of Thoughts์—์„œ๋Š” ๋‹ค์„ฏ ๊ฐ€์ง€ ์œ ํ˜•์˜ ์›์ž๋ฅผ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค:

  1. premise (์ „์ œ): ๋ฌธ์ œ ํ•ด๊ฒฐ์„ ์œ„ํ•œ ๊ธฐ๋ณธ ๊ฐ€์ •์ด๋‚˜ ์ฃผ์–ด์ง„ ์ •๋ณด
  2. reasoning (์ถ”๋ก ): ๋‹ค๋ฅธ ์›์ž๋“ค์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ ๋…ผ๋ฆฌ์  ์ถ”๋ก  ๊ณผ์ •
  3. hypothesis (๊ฐ€์„ค): ๊ฐ€๋Šฅํ•œ ํ•ด๊ฒฐ์ฑ…์ด๋‚˜ ์ค‘๊ฐ„ ๊ฒฐ๋ก ์— ๋Œ€ํ•œ ์ œ์•ˆ
  4. verification (๊ฒ€์ฆ): ๋‹ค๋ฅธ ์›์ž(ํŠนํžˆ ๊ฐ€์„ค)์˜ ์œ ํšจ์„ฑ์„ ํ‰๊ฐ€ํ•˜๋Š” ๊ณผ์ •
  5. conclusion (๊ฒฐ๋ก ): ๊ฒ€์ฆ๋œ ๊ฐ€์„ค์ด๋‚˜ ์ตœ์ข… ๋ฌธ์ œ ํ•ด๊ฒฐ์ฑ…

ํ•ต์‹ฌ ๊ธฐ๋Šฅ

1. ๋ถ„ํ•ด-์ˆ˜์ถ• ๋ฉ”์ปค๋‹ˆ์ฆ˜ (Decomposition-Contraction)

์›์ž๋ฅผ ๋” ์ž‘์€ ํ•˜์œ„ ์›์ž๋กœ ๋ถ„ํ•ดํ•˜๊ณ  ๊ฒ€์ฆ ํ›„ ๋‹ค์‹œ ์ˆ˜์ถ•ํ•˜๋Š” ๋ฉ”์ปค๋‹ˆ์ฆ˜์ž…๋‹ˆ๋‹ค.

  • ์›์ž ๋ถ„ํ•ด (Decomposition): ๋ณต์žกํ•œ ์›์ž๋ฅผ ๋” ์ž‘์€ ํ•˜์œ„ ์›์ž๋กœ ๋ถ„ํ•ดํ•ฉ๋‹ˆ๋‹ค.

    • startDecomposition(atomId): ์›์ž ๋ถ„ํ•ด ์‹œ์ž‘
    • addToDecomposition(decompositionId, atomId): ๋ถ„ํ•ด์— ํ•˜์œ„ ์›์ž ์ถ”๊ฐ€
    • completeDecomposition(decompositionId): ๋ถ„ํ•ด ๊ณผ์ • ์™„๋ฃŒ
  • ์›์ž ์ˆ˜์ถ• (Contraction): ํ•˜์œ„ ์›์ž๋“ค์ด ๋ชจ๋‘ ๊ฒ€์ฆ๋˜๋ฉด ์›๋ž˜ ์›์ž๋กœ ๋‹ค์‹œ ์ˆ˜์ถ•ํ•ฉ๋‹ˆ๋‹ค.

    • ํ•˜์œ„ ์›์ž๋“ค์˜ ์‹ ๋ขฐ๋„์— ๊ธฐ๋ฐ˜ํ•˜์—ฌ ์›๋ž˜ ์›์ž์˜ ์‹ ๋ขฐ๋„๋ฅผ ๊ณ„์‚ฐ
    • ๊ฒ€์ฆ๋œ ๊ฐ€์„ค์ด ๊ณ ์‹ ๋ขฐ๋„๋ฅผ ๊ฐ€์ง€๋ฉด ์ž๋™์œผ๋กœ ๊ฒฐ๋ก ์„ ์ œ์•ˆ
2. ์ž๋™ ์ข…๋ฃŒ ๋ฉ”์ปค๋‹ˆ์ฆ˜ (Automatic Termination)
  • ์ตœ๋Œ€ ๊นŠ์ด(depth)์— ๋„๋‹ฌํ•˜๊ฑฐ๋‚˜ ๋†’์€ ์‹ ๋ขฐ๋„์˜ ๊ฒฐ๋ก ์„ ์ฐพ์œผ๋ฉด ์ž๋™ ์ข…๋ฃŒ๋ฉ๋‹ˆ๋‹ค.
  • getTerminationStatus(): ํ˜„์žฌ ์ข…๋ฃŒ ์ƒํƒœ์™€ ์ด์œ ๋ฅผ ๋ฐ˜ํ™˜
  • getBestConclusion(): ๊ฐ€์žฅ ๋†’์€ ์‹ ๋ขฐ๋„์˜ ๊ฒฐ๋ก ์„ ๋ฐ˜ํ™˜

๋งค๊ฐœ๋ณ€์ˆ˜ ์„ค๋ช…

  • atomId: ์›์ž์˜ ๊ณ ์œ  ์‹๋ณ„์ž (์˜ˆ: 'A1', 'H2' ๋“ฑ)
  • content: ์›์ž์˜ ์‹ค์ œ ๋‚ด์šฉ
  • atomType: ์›์ž์˜ ์œ ํ˜• (premise, reasoning, hypothesis, verification, conclusion ์ค‘ ํ•˜๋‚˜)
  • dependencies: ์ด ์›์ž๊ฐ€ ์˜์กดํ•˜๋Š” ๋‹ค๋ฅธ ์›์ž๋“ค์˜ ID ๋ชฉ๋ก
  • confidence: ์ด ์›์ž์˜ ์‹ ๋ขฐ๋„ (0~1 ์‚ฌ์ด์˜ ๊ฐ’)
  • isVerified: ์ด ์›์ž๊ฐ€ ๊ฒ€์ฆ๋˜์—ˆ๋Š”์ง€ ์—ฌ๋ถ€
  • depth: ์ด ์›์ž์˜ ๊นŠ์ด (๋ถ„ํ•ด-์ˆ˜์ถ• ํ”„๋กœ์„ธ์Šค์—์„œ์˜ ๊นŠ์ด ์ˆ˜์ค€)

์‚ฌ์šฉ ๋ฐฉ๋ฒ•

  1. ๋ฌธ์ œ๋ฅผ ์ดํ•ดํ•˜๊ณ  ํ•„์š”ํ•œ ์ „์ œ(premise) ์›์ž๋“ค์„ ์ •์˜
  2. ์ „์ œ๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ์ถ”๋ก (reasoning) ์›์ž ์ƒ์„ฑ
  3. ์ถ”๋ก ์„ ๋ฐ”ํƒ•์œผ๋กœ ๊ฐ€์„ค(hypothesis) ์›์ž ์ƒ์„ฑ
  4. ๊ฐ€์„ค์„ ๊ฒ€์ฆ(verification)ํ•˜๋Š” ์›์ž ์ƒ์„ฑ
  5. ๊ฒ€์ฆ๋œ ๊ฐ€์„ค์„ ๋ฐ”ํƒ•์œผ๋กœ ๊ฒฐ๋ก (conclusion) ์›์ž ๋„์ถœ
  6. ํ•„์š”์‹œ ์›์ž ๋ถ„ํ•ด(decomposition)๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ๋” ๊นŠ์ด ํƒ์ƒ‰
  7. ๋†’์€ ์‹ ๋ขฐ๋„์˜ ๊ฒฐ๋ก  ์›์ž๋ฅผ ์ตœ์ข… ๋‹ต๋ณ€์œผ๋กœ ์ œ์‹œ

Sequential Thinking๊ณผ Atom of Thoughts ๋น„๊ต (์กฐ๊ธˆ ๋” ํ…Œ์ŠคํŠธ๊ฐ€ ํ•„์š”ํ•จ)

๋‘ ๊ฐ€์ง€ ์‚ฌ๊ณ  ๋„๊ตฌ๋ฅผ ๋™์ผํ•œ ์ฃผ์ œ์— ์ ์šฉํ•œ ํ›„ ๊ด€์ฐฐ๋œ ์ฐจ์ด์ ๊ณผ ์„ฑ๋Šฅ ํŠน์„ฑ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์Šต๋‹ˆ๋‹ค:

๊ตฌ์กฐ์  ์ฐจ์ด์ 

Sequential Thinking:

  • ์„ ํ˜•์  ์‚ฌ๊ณ  ๊ณผ์ •: ํ•œ ์‚ฌ๊ณ ์—์„œ ๋‹ค์Œ ์‚ฌ๊ณ ๋กœ ์ˆœ์ฐจ์ ์œผ๋กœ ์ง„ํ–‰
  • ์ „์ฒด ์‚ฌ๊ณ  ์ˆ˜๋ฅผ ๋ฏธ๋ฆฌ ์˜ˆ์ธก
  • ๊ฐ ์‚ฌ๊ณ  ๋‹จ๊ณ„๋Š” ์ด์ „ ๋‹จ๊ณ„๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋จ

Atom of Thoughts:

  • ๋น„์„ ํ˜•, ๋„คํŠธ์›Œํฌ ๊ตฌ์กฐ: ์—ฌ๋Ÿฌ ์‚ฌ๊ณ  ๋‹จ์œ„(์›์ž)๊ฐ€ ์˜์กด์„ฑ์„ ๊ฐ€์ง€๊ณ  ์—ฐ๊ฒฐ๋จ
  • ์›์ž ์œ ํ˜•(์ „์ œ, ์ถ”๋ก , ๊ฐ€์„ค, ๊ฒ€์ฆ, ๊ฒฐ๋ก )์— ๋”ฐ๋ผ ์ฒด๊ณ„์ ์ธ ๊ตฌ์กฐ ํ˜•์„ฑ
  • ๊ฐ ์›์ž์˜ ์‹ ๋ขฐ๋„ ์ˆ˜์ค€์„ ๋ช…์‹œ์ ์œผ๋กœ ํ‰๊ฐ€
๋น„๊ต ๊ฐ•์ 

Sequential Thinking ๊ฐ•์ :

  • ์ง๊ด€์  ํ๋ฆ„: ์ž์—ฐ์Šค๋Ÿฌ์šด ์ธ๊ฐ„์˜ ์‚ฌ๊ณ  ๊ณผ์ •๊ณผ ์œ ์‚ฌ
  • ๋‹จ์ˆœ์„ฑ: ๊ฐ„๋‹จํ•œ ๊ตฌ์กฐ๋กœ ์ง๊ด€์ ์ธ ๋ฌธ์ œ์— ๋น ๋ฅด๊ฒŒ ์ ์šฉ ๊ฐ€๋Šฅ
  • ์œ ์—ฐ์„ฑ: ์‚ฌ๊ณ  ๊ณผ์ • ์ค‘์— ์ด์ „ ๋‹จ๊ณ„๋ฅผ ์ˆ˜์ •ํ•˜๊ฑฐ๋‚˜ ๋ฐฉํ–ฅ์„ ๋ณ€๊ฒฝํ•  ์ˆ˜ ์žˆ์Œ

Atom of Thoughts ๊ฐ•์ :

  • ์‹ ๋ขฐ๋„ ํ‰๊ฐ€: ๊ฐ ์‚ฌ๊ณ ์˜ ์‹ ๋ขฐ๋„๋ฅผ ๋ช…์‹œ์ ์œผ๋กœ ์ธก์ •ํ•˜์—ฌ ๊ฒฐ๋ก ์˜ ์œ ํšจ์„ฑ ๊ฐœ์„ 
  • ๊ฒ€์ฆ ๊ณผ์ •: ์ฒด๊ณ„์ ์ธ ๊ฒ€์ฆ ๋‹จ๊ณ„๋ฅผ ํ†ตํ•ด ๊ฐ€์„ค ํ‰๊ฐ€
  • ์˜์กด์„ฑ ์ถ”์ : ์–ด๋–ค ์ „์ œ๋‚˜ ์ถ”๋ก ์ด ํŠน์ • ๊ฒฐ๋ก ์— ์˜ํ–ฅ์„ ๋ฏธ์ณค๋Š”์ง€ ๋ช…ํ™•ํ•˜๊ฒŒ ์ถ”์ 
  • ๋ณ‘๋ ฌ ์ฒ˜๋ฆฌ: ์—ฌ๋Ÿฌ ์‚ฌ๊ณ  ์›์ž๋ฅผ ๋™์‹œ์— ๊ณ ๋ ค ๊ฐ€๋Šฅ
ํšจ์œจ์„ฑ๊ณผ ์ •ํ™•์„ฑ

ํšจ์œจ์„ฑ:

  • Sequential Thinking: ๋‹จ์ˆœํ•œ ๋ฌธ์ œ์— ๋” ํšจ์œจ์ ์ด๋ฉฐ, ์‚ฌ๊ณ ๊ฐ€ ๋น ๋ฅด๊ฒŒ ์ง„ํ–‰๋จ
  • Atom of Thoughts: ๋ณต์žกํ•œ ๋ฌธ์ œ์— ๋” ํšจ์œจ์ ์ด์ง€๋งŒ, ์ฒด๊ณ„์ ์ธ ๊ตฌ์กฐ๋ฅผ ๋งŒ๋“œ๋Š” ์ดˆ๊ธฐ ์˜ค๋ฒ„ํ—ค๋“œ๊ฐ€ ์žˆ์Œ

์ •ํ™•์„ฑ:

  • Sequential Thinking: ์‚ฌ๊ณ  ๊ณผ์ •์ด ๊นŠ์–ด์งˆ์ˆ˜๋ก ์ด์ „ ๋‹จ๊ณ„์—์„œ์˜ ์˜ค๋ฅ˜ ๋ˆ„์  ๊ฐ€๋Šฅ์„ฑ
  • Atom of Thoughts: ๊ฒ€์ฆ ๋‹จ๊ณ„์™€ ์‹ ๋ขฐ๋„ ํ‰๊ฐ€๋ฅผ ํ†ตํ•ด ์˜ค๋ฅ˜ ๊ฐ€๋Šฅ์„ฑ ๊ฐ์†Œ, ๋” ์‹ ๋ขฐํ•  ์ˆ˜ ์žˆ๋Š” ๊ฒฐ๋ก  ๋„์ถœ
๋ชฉ์ ๋ณ„ ์ ํ•ฉ์„ฑ

Sequential Thinking์— ์ ํ•ฉํ•œ ๊ฒฝ์šฐ:

  • ๋‹จ์ˆœํ•˜๊ฑฐ๋‚˜ ์ค‘๊ฐ„ ์ •๋„ ๋ณต์žกํ•œ ๋ฌธ์ œ
  • ์‹œ๊ฐ„ ์ œ์•ฝ์ด ์žˆ๋Š” ์ƒํ™ฉ
  • ์ž์—ฐ์Šค๋Ÿฌ์šด ์Šคํ† ๋ฆฌํ…”๋ง์ด๋‚˜ ์„ค๋ช…์ด ํ•„์š”ํ•œ ๊ฒฝ์šฐ

Atom of Thoughts์— ์ ํ•ฉํ•œ ๊ฒฝ์šฐ:

  • ๋งค์šฐ ๋ณต์žกํ•œ ๋ฌธ์ œ
  • ์ •ํ™•์„ฑ๊ณผ ์‹ ๋ขฐ์„ฑ์ด ์ค‘์š”ํ•œ ์ƒํ™ฉ
  • ์—ฌ๋Ÿฌ ๊ด€์ ์—์„œ ๊ฒ€์ฆ์ด ํ•„์š”ํ•œ ๊ฐ€์„ค
  • ๋ณต์žกํ•œ ์˜์กด ๊ด€๊ณ„๊ฐ€ ์žˆ๋Š” ์ถ”๋ก 
๊ฒฐ๋ก 

๋‘ ๋„๊ตฌ ๋ชจ๋‘ ์ธ๊ณต ์ง€๋Šฅ์˜ ์ถ”๋ก  ๋Šฅ๋ ฅ์„ ํ–ฅ์ƒ์‹œํ‚ค๋Š” ๋ฐ ๊ธฐ์—ฌํ•  ์ˆ˜ ์žˆ์ง€๋งŒ, ์ ์ ˆํ•œ ๋„๊ตฌ๋Š” ๋ฌธ์ œ์˜ ํŠน์„ฑ๊ณผ ์š”๊ตฌ ์‚ฌํ•ญ์— ๋”ฐ๋ผ ๋‹ฌ๋ผ์ง‘๋‹ˆ๋‹ค. Sequential Thinking์€ ์ง๊ด€์ ์ด๊ณ  ๋น ๋ฅธ ์‚ฌ๊ณ  ๊ณผ์ •์ด ํ•„์š”ํ•  ๋•Œ ์œ ์šฉํ•˜๋ฉฐ, Atom of Thoughts๋Š” ์ฒด๊ณ„์ ์ธ ๊ฒ€์ฆ๊ณผ ๋†’์€ ์‹ ๋ขฐ์„ฑ์ด ํ•„์š”ํ•œ ๋ณต์žกํ•œ ๋ฌธ์ œ์— ๋” ์ ํ•ฉํ•ฉ๋‹ˆ๋‹ค.

๋ช…๋ น์–ด ๋„๊ตฌ (atomcommands)

Atom of Thoughts์˜ ๋ถ„ํ•ด-์ˆ˜์ถ• ๋ฉ”์ปค๋‹ˆ์ฆ˜๊ณผ ์ž๋™ ์ข…๋ฃŒ๋ฅผ ์ œ์–ดํ•˜๋Š” ๋ช…๋ น์–ด ๋„๊ตฌ์ž…๋‹ˆ๋‹ค.

์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ๋ช…๋ น์–ด:

  1. decompose: ์ง€์ •๋œ ์›์ž๋ฅผ ๋” ์ž‘์€ ํ•˜์œ„ ์›์ž๋กœ ๋ถ„ํ•ดํ•ฉ๋‹ˆ๋‹ค.
    • ํ•„์š” ๋งค๊ฐœ๋ณ€์ˆ˜: atomId
  2. complete_decomposition: ์ง„ํ–‰ ์ค‘์ธ ๋ถ„ํ•ด ํ”„๋กœ์„ธ์Šค๋ฅผ ์™„๋ฃŒํ•ฉ๋‹ˆ๋‹ค.
    • ํ•„์š” ๋งค๊ฐœ๋ณ€์ˆ˜: decompositionId
  3. termination_status: ํ˜„์žฌ AoT ํ”„๋กœ์„ธ์Šค์˜ ์ข…๋ฃŒ ์ƒํƒœ๋ฅผ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.
  4. best_conclusion: ๊ฐ€์žฅ ๋†’์€ ์‹ ๋ขฐ๋„์˜ ๊ฒ€์ฆ๋œ ๊ฒฐ๋ก ์„ ๊ฐ€์ ธ์˜ต๋‹ˆ๋‹ค.
  5. set_max_depth: ์ตœ๋Œ€ ๊นŠ์ด ์ œํ•œ์„ ๋ณ€๊ฒฝํ•ฉ๋‹ˆ๋‹ค.
    • ํ•„์š” ๋งค๊ฐœ๋ณ€์ˆ˜: maxDepth

Installing via Smithery

To install Atom of Thoughts for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @kbsooo/mcp_atom_of_thoughts --client claude

MCP ์„œ๋ฒ„ ์„ค์ • ๋ฐฉ๋ฒ•

Atom of Thoughts MCP ์„œ๋ฒ„๋ฅผ ์‚ฌ์šฉํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” Claude Desktop ๋˜๋Š” Cline์˜ MCP ์„ค์ •์— ์„œ๋ฒ„๋ฅผ ๋“ฑ๋กํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๋‹ค์Œ์€ ์„œ๋ฒ„ ๊ตฌ์„ฑ์˜ ์˜ˆ์‹œ์ž…๋‹ˆ๋‹ค:

{ 
  "mcpServers": { 
    "atom-of-thoughts": { 
      "command": "node", 
      "args": ["/ABSOLUTE/PATH/TO/PARENT/FOLDER/atom-of-thoughts/build/index.js"], 
      "disabled": false, 
      "autoApprove": [] 
    } 
  } 
}

์œ„ ์„ค์ •์—์„œ /ABSOLUTE/PATH/TO/PARENT/FOLDER๋Š” ์‹ค์ œ ํ”„๋กœ์ ํŠธ๊ฐ€ ์œ„์น˜ํ•œ ์ ˆ๋Œ€ ๊ฒฝ๋กœ๋กœ ๋ณ€๊ฒฝํ•ด์ฃผ์„ธ์š”. ์„ค์ •์„ ์ €์žฅํ•œ ํ›„ Claude Desktop ๋˜๋Š” Cline์„ ์žฌ์‹œ์ž‘ํ•˜๋ฉด Atom of Thoughts MCP ์„œ๋ฒ„๋ฅผ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.