⚠️ 중요: 프로젝트 상태 ⚠️
이 프로젝트는 초기 개발 단계(v0.1.0)이며 아직 정식 출시 준비가 되지 않았습니다 . 현재 몇 가지 중요한 문제를 해결하기 위해 대대적인 재작업이 진행 중입니다. 현재 제한 사항 및 계획된 개선 사항은 문제 페이지를 확인하세요.
중요한 워크플로에 이 기능을 사용하기 전에 안정적인 릴리스(v0.2.0+)가 나올 때까지 기다리는 것이 좋습니다.
클로드에게 정교한 컨텍스트 관리 기능을 제공하는 MCP(모델 컨텍스트 프로토콜) 서버로, 세션 전반에 걸쳐 지속적인 컨텍스트, 프로젝트별 컨텍스트 구성 및 대화 연속성을 지원합니다.
현재 서버에는 Claude Desktop 이외의 MCP 클라이언트와 호환성 문제가 있습니다.
특정 프로젝트 ID가 없으면 컨텍스트 목록 기능이 제한됩니다.
보안 기능은 최소한이며 생산에 적합하지 않습니다.
오류 처리가 기본이며 도움이 되는 지침을 제공하지 못할 수 있습니다.
테스트 인프라가 구축되어 있지 않습니다.
Related MCP server: MCP Memory Keeper
이 프로젝트는 현재 활발하게 개선되고 있습니다. 향후 주요 개선 사항은 다음과 같습니다.
안정성 개선 - 홈 디렉토리 확인 및 컨텍스트 목록의 핵심 문제 해결
향상된 오류 처리 - 더 나은 오류 메시지 및 복구 메커니즘
보안 강화 - 입력 검증, 경로 정리 및 데이터 보호
고급 컨텍스트 관리 - 버전 관리, 검색 및 더 나은 구성
더 자세한 로드맵을 보려면 종합 분석 섹션을 참조하세요.
프로젝트 컨텍스트 관리
계층적 맥락 구성
부모-자녀 관계
컨텍스트 간 교차 참조
프로젝트별 메타데이터
대화의 연속성
세션 기반 컨텍스트 추적
대화 연쇄
메타데이터가 풍부한 컨텍스트 저장소
유연한 태깅 시스템
효율적인 보관
체계적인 디렉토리 구조
JSON 기반 저장소
빠른 조회 인덱싱
비동기 작업
서버는 Claude 데스크톱 앱의 MCP 설정에서 자동으로 구성됩니다. 모든 컨텍스트는 더 나은 정리를 위해 ~/.claude/ 에 저장됩니다.
지엑스피1
// Save project context
use_mcp_tool({
server_name: "claude-server",
tool_name: "save_project_context",
arguments: {
id: "feature-design-v1",
projectId: "my-project",
content: "Design discussion...",
parentContextId: "requirements-v1",
references: ["api-spec-v1"],
tags: ["design"],
metadata: { status: "in-progress" }
}
});// Save conversation context
use_mcp_tool({
server_name: "claude-server",
tool_name: "save_conversation_context",
arguments: {
id: "chat-2024-01-01",
sessionId: "session-123",
content: "Discussion content...",
continuationOf: "previous-chat-id",
tags: ["meeting"]
}
});// Get context
use_mcp_tool({
server_name: "claude-server",
tool_name: "get_context",
arguments: {
id: "feature-design-v1",
projectId: "my-project"
}
});
// List contexts
use_mcp_tool({
server_name: "claude-server",
tool_name: "list_contexts",
arguments: {
projectId: "my-project",
tag: "design",
type: "project"
}
});컨텍스트 관리 가이드 - 컨텍스트 유형 및 사용에 대한 자세한 가이드
아키텍처 개요 - 기술 구현 세부 정보
사용 가이드 - 일반 사용 지침
Claude Desktop 통합 - Claude Desktop과 통합
저장소를 복제합니다
종속성 설치:
npm install서버를 빌드하세요:
npm run build서버는 build/index.js 로 빌드됩니다.
서버는 ~/Library/Application Support/Claude/claude_desktop_config.json 에 있는 Claude 데스크톱 앱의 구성 파일을 통해 구성됩니다.
{
"mcpServers": {
"claude-server": {
"command": "node",
"args": ["/path/to/claude-server/build/index.js"]
}
}
}기여를 환영합니다! 이슈를 제출하거나 풀 리퀘스트를 보내주세요.
MIT
Retrieve context by ID and optional project ID
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | ID of the context to retrieve | |
| projectId | No | Optional project ID for project contexts |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It states the action ('retrieve') but lacks behavioral details such as whether this is a read-only operation, error handling, permissions needed, or rate limits. The description is minimal and doesn't 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 extremely concise with a single sentence that directly states the tool's purpose. It's front-loaded and wastes no words, making it efficient for quick understanding.
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 no annotations and no output schema, the description is incomplete. It doesn't explain what 'context' entails, the return format, or how it interacts with sibling tools. For a tool with 2 parameters and behavioral uncertainty, more context is needed to be fully helpful.
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 fully documents both parameters. The description adds no additional meaning beyond what's in the schema (e.g., it doesn't explain context types or project relationships). Baseline 3 is appropriate as the schema handles parameter documentation.
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 verb ('retrieve') and resource ('context'), specifying it's done by ID with an optional project ID. However, it doesn't differentiate from sibling tools like 'list_contexts' or 'save_conversation_context', which would require more specific scope or purpose details.
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?
No guidance is provided on when to use this tool versus alternatives. The description mentions an optional project ID but doesn't explain when to include it or how this tool differs from siblings like 'list_contexts' or save operations.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List contexts with filtering options
| Name | Required | Description | Default |
|---|---|---|---|
| projectId | No | Optional project ID to filter by | |
| tag | No | Optional tag to filter by | |
| type | No | Optional type to filter by |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries full burden but offers minimal behavioral insight. It implies a read operation with filtering but doesn't disclose critical details like pagination, rate limits, authentication needs, or what 'list contexts' entails (e.g., format, scope). This leaves significant gaps for agent understanding.
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 extremely concise with a single sentence that front-loads the core purpose ('List contexts') and adds a brief qualifier ('with filtering options'). There is zero wasted verbiage, making it highly efficient.
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 no annotations, no output schema, and a read operation with filtering, the description is incomplete. It lacks details on return values, error handling, or behavioral constraints, leaving the agent with insufficient context to use the tool effectively beyond basic parameter input.
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 fully documents all three optional parameters (projectId, tag, type with enum). The description adds no additional meaning beyond implying filtering exists, matching the baseline for high schema coverage without extra param insights.
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 action ('List contexts') and mentions filtering capabilities, which distinguishes it from simple listing operations. However, it doesn't explicitly differentiate from sibling tools like 'get_context' (which might retrieve a single context) or the save operations, missing full sibling differentiation.
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 like 'get_context' for single context retrieval or the save tools for creation. It mentions filtering options but doesn't specify scenarios or prerequisites for usage.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Save conversation context with continuation support
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | Unique identifier for the context | |
| sessionId | Yes | Conversation session identifier | |
| content | Yes | Context content to save | |
| continuationOf | No | Optional ID of previous context | |
| tags | No | Optional tags for categorizing | |
| metadata | No | Optional additional metadata |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries full burden for behavioral disclosure. 'Save' implies a write operation, but the description doesn't address permissions needed, whether this overwrites existing context with the same ID, what happens on success/failure, or any rate limits. The 'continuation support' hint is useful but insufficient for a mutation tool with zero annotation coverage.
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 extremely concise at just 5 words, front-loading the core purpose. Every word earns its place: 'Save' (action), 'conversation context' (resource), 'with continuation support' (key feature). There's zero waste or redundancy.
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?
For a mutation tool with 6 parameters, no annotations, and no output schema, the description is incomplete. It doesn't explain what happens after saving, what format the saved context takes, whether there are size limits on content, or how continuation actually works. The agent lacks crucial information about this write operation's behavior and outcomes.
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 all parameters are documented in the schema itself. The description adds no additional parameter semantics beyond what's already in the schema descriptions. The baseline score of 3 is appropriate when the schema does the heavy lifting, though the description could have explained relationships between parameters like how 'continuationOf' relates to 'id'.
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 states the tool saves conversation context with continuation support, which is a clear verb+resource combination. However, it doesn't distinguish this from its sibling 'save_project_context' - both appear to save context but for different types (conversation vs project). The purpose is understandable but lacks sibling differentiation.
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. There's no mention of when to choose 'save_conversation_context' over 'save_project_context', nor any prerequisites or typical use cases. The agent must infer usage from the tool name alone.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Save project-specific context with relationships
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | Unique identifier for the context | |
| projectId | Yes | Project identifier | |
| content | Yes | Context content to save | |
| parentContextId | No | Optional ID of parent context | |
| references | No | Optional related context IDs | |
| tags | No | Optional tags for categorizing | |
| metadata | No | Optional additional metadata |
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 the tool saves context with relationships, implying a write operation, but doesn't disclose critical behaviors like whether it overwrites existing context with the same ID, what permissions are required, error conditions, or response format. For a mutation tool with zero annotation coverage, this is a significant gap.
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 a single, efficient sentence that front-loads the core purpose ('save project-specific context with relationships') with zero waste. Every word earns its place, making it 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's complexity (7 parameters, mutation operation, no output schema, and no annotations), the description is incomplete. It doesn't explain the return values, error handling, or behavioral nuances like how 'relationships' are enforced or what happens on duplicate IDs. For a save operation with rich parameters, more context is needed.
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 7 parameters thoroughly. The description adds no additional meaning beyond implying relationships via 'parentContextId' and 'references', which is already clear from the schema. With high schema coverage, the baseline is 3, and the description doesn't compensate with extra insights.
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 action ('save') and resource ('project-specific context with relationships'), which is specific and actionable. It distinguishes from sibling 'save_conversation_context' by specifying 'project-specific' context, though it doesn't explicitly differentiate from 'get_context' or 'list_contexts' beyond the save vs. get/list distinction.
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 like 'save_conversation_context' for conversation contexts or when to retrieve vs. save using 'get_context'/'list_contexts'. It lacks explicit when/when-not instructions or prerequisites, leaving usage context implied by the tool name alone.
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.
v1.0.0get_contextlist_contextssave_conversation_contextsave_project_contextScored across 4 tools
Each tool has a clearly distinct purpose with no overlap: get_context retrieves a specific context, list_contexts lists contexts with filters, save_conversation_context saves conversation-specific context, and save_project_context saves project-specific context. The descriptions clearly differentiate between retrieval, listing, and two types of saving operations.
All tools follow a consistent verb_noun pattern (get_context, list_contexts, save_conversation_context, save_project_context) with clear, descriptive names. The naming convention is uniform throughout the set, making it easy to understand each tool's function at a glance.
With 4 tools, the count is reasonable for a context management server, covering key operations like retrieval, listing, and saving. It feels slightly minimal but well-scoped, as each tool serves a distinct and necessary function without redundancy.
The tool set provides good coverage for context management with get, list, and save operations for both conversation and project contexts. A minor gap exists in update or delete functionality for contexts, but agents can likely work around this by re-saving or managing contexts through the provided tools.
Persistent context for Claude. Your AI always knows your projects and next actions across sessions.
Share one project context across ChatGPT, Claude, Telegram and any MCP client.
Persistent memory for Claude Code and Cursor. Stop re-explaining your project every session.
Cross-LLM persistent memory: store context once, recall it from any AI model.
