kothar
kothar
上下文感知的 MCP 服务器顾问。告诉您针对特定项目需要安装什么，以及原因。
问题所在
Glama 拥有超过 19,000 个 MCP 服务器。您有一个项目。但没有人能弥合两者之间的差距。
直接询问 LLM 会导致它们幻觉出不存在的服务器，并根据过时的训练数据进行推荐。目录只能提供搜索，不能提供建议。
kothar 填补了上下文选择的空白：不是“这里有 19,000 个选项”，而是“针对您当前的项目，此时此刻，您需要什么以及为什么需要”。
两个无人服务的时刻
项目启动： “我正在使用 DuckDB 和 FastAPI 构建一个 Python 数据管道” → 我现在应该安装什么
项目中期： “我刚刚添加了一个身份验证层 / 我需要处理 PDF 摄取” → 既然我已经到了这一步，现在应该添加什么
第二个时刻更有价值。在项目启动时，人们可以搜索 Google。而在项目中期，他们正处于工作流中。
三个工具
recommend_for_project(description)
→ top MCP servers for your stack with rationale
recommend_next(current_stack, new_context)
→ what to add as your project evolves
explain_why(server_name, project_description)
→ why a specific server fits your project安装
先决条件： uv
git clone https://github.com/yahiaklk/kothar
cd kothar
uv sync构建索引（首次运行，约 30 秒）：
uv run python -m kothar.indexer添加到 Claude Code
claude mcp add --scope user kothar -- uv run --directory /path/to/kothar python -m kothar.server添加到 Claude Desktop
添加到您的 claude_desktop_config.json：
{
"mcpServers": {
"kothar": {
"command": "uv",
"args": ["run", "--directory", "/path/to/kothar", "python", "-m", "kothar.server"]
}
}
}使用方法
连接后，询问您的 AI 助手：
recommend_for_project("Python FastAPI backend with PostgreSQL and JWT auth")
recommend_next("github,filesystem", "adding Stripe payments and PDF invoices")
explain_why("postgres", "multi-tenant SaaS with row-level security")工作原理
解析 awesome-mcp-servers（2000 多个精选服务器）
使用
all-MiniLM-L6-v2对描述进行嵌入（本地，无 API 成本）存储在 DuckDB 中，使用余弦相似度进行查询
基于模板的逻辑说明 — 基于注册表，而非幻觉
重建索引
uv run python -m kothar.indexer --forceDocker
多阶段镜像，内置嵌入模型 + DuckDB 索引 — 无运行时网络依赖。
docker build -t kothar:0.3.0 .
docker run --rm -i kothar:0.3.0 # stdio transport, for local MCP clients接入 Claude Desktop：
{
"mcpServers": {
"kothar": {
"command": "docker",
"args": ["run", "--rm", "-i", "kothar:0.3.0"]
}
}
}非 root 用户 (uid=10001)，固定 Python 3.12，依赖项从 uv.lock 解析，模型在运行时缓存于 /app/.hf_cache，并设置 HF_HUB_OFFLINE=1。
技术栈
Python · FastMCP · DuckDB · sentence-transformers · uv
许可证
Available Tools
4 toolsexplain_whyB
Explain why a specific MCP server is a good fit for a given project. Example: server_name="github", project_description="open source Python library with CI/CD"
| Name | Required | Description | Default |
|---|---|---|---|
| server_name | Yes | ||
| project_description | Yes |
Output Schema
| Name | Required | Description |
|---|---|---|
| result | Yes |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations and description only states purpose. Does not disclose behavior like error handling, required permissions, or output format beyond what output schema might provide.
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?
Description is a single sentence plus example, which is concise. However, the structure could be improved by adding a brief usage context.
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 low complexity and presence of output schema, description is minimally adequate but lacks usage guidelines and behavioral details that would make it fully self-contained.
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 coverage is 0%, meaning parameters have no descriptions. The description provides an example usage but does not explain parameter semantics beyond the example.
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?
Describes a clear verb+resource combination: 'Explain why a specific MCP server is a good fit for a given project.' This distinguishes it from sibling recommendation tools.
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?
Provides an example but no explicit guidance on when to use this tool versus the sibling recommendation tools. The example helps but does not fully clarify context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
recommend_for_goalA
Decompose a multi-part goal into sub-queries and recommend MCP servers for each part. Splits on hard boundaries: '. ', '; ', ' then ', ', then ', ' and then ' (not bare ' and '). project: optional project context prepended to each sub-query for richer semantic matching. Example: goal="integrate GitHub. add Stripe payments", project="Python FastAPI backend"
| Name | Required | Description | Default |
|---|---|---|---|
| goal | Yes | ||
| project | No |
Output Schema
| Name | Required | Description |
|---|---|---|
| result | Yes |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided, so description carries the burden. It explains splitting behavior and project prepending but does not disclose whether the tool is read-only or destructive, side effects, or error handling. Adequate but lacks safety profile.
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?
Three sentences plus example, no wasted words. Front-loaded main purpose, then splitting details, then optional parameter. Could be more structured but 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?
Output schema exists, so return values not required. Explains input and behavior well, but lacks usage guidelines for sibling differentiation and does not mention prerequisites or error conditions. Fairly complete for a tool with output schema.
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 0%, yet the description adds meaning: 'goal' is a multi-part goal with splitting rules, 'project' is optional context prepended. This goes beyond the schema's minimal type info.
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?
Clearly states the tool decomposes multi-part goals into sub-queries and recommends MCP servers. Distinguishes from siblings like recommend_next and recommend_for_project by focusing on multi-part goals and splitting logic.
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?
Provides splitting rules and optional parameter description but does not explicitly state when to use this tool versus alternatives like recommend_for_project or recommend_next. Usage is implied for multi-part goals.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
recommend_for_projectA
Given a project description, recommend the top MCP servers to install and explain why each one fits. Example: "Python FastAPI backend with PostgreSQL and JWT auth"
| Name | Required | Description | Default |
|---|---|---|---|
| description | Yes |
Output Schema
| Name | Required | Description |
|---|---|---|
| result | Yes |
TDQS
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 implies a read-like recommendation behavior but does not disclose details like whether it modifies state, requires authentication, or has limitations on input length.
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 two sentences and an example, front-loading the core purpose. Every sentence adds value.
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 existence of an output schema, the description does not need to detail return values. It covers what the tool does and gives an example, but could mention that it returns a list of servers with explanations. The hint 'top' could be clarified.
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% coverage, so the description must add meaning. It specifies that the parameter is a 'project description' and provides an example, which clarifies the expected format beyond the bare schema.
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 function: given a project description, it recommends MCP servers and explains why. It distinguishes from siblings like 'recommend_for_goal' by specifying the input is a project description, not a goal.
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 context for when to use (when you have a project description) with an example, but does not mention when not to use or explicitly contrast with sibling tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
recommend_nextA
Mid-project advisor: given your current MCP stack (list of server names) and a new development context, recommend what to add next and why. Example: current_stack=["github", "filesystem"], new_context="adding Stripe payments and PDF invoices" session_file: optional path to a session notes file whose content is appended to new_context.
| Name | Required | Description | Default |
|---|---|---|---|
| current_stack | Yes | ||
| new_context | Yes | ||
| session_file | No |
Output Schema
| Name | Required | Description |
|---|---|---|
| result | Yes |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full transparency burden. It explains the tool recommends with reasoning and mentions an optional session_file. The output schema exists but is not shown; the description does not cover output format, but the example hints at a recommendation with reasons. Overall, it is sufficiently transparent for a recommendation tool.
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 plus an example, front-loaded with purpose, and contains no unnecessary words. It is efficiently structured and easy to parse.
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 recommendation nature and the presence of an output schema (which handles return value documentation), the description covers all essential input semantics and usage context with an example. It is fully complete for agent decision-making and invocation.
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 0%, so the description must compensate. It explains current_stack as 'list of server names', new_context as 'development context', and session_file as 'optional path to session notes file whose content is appended to new_context'. This adds meaningful context beyond the schema. A slight improvement could clarify the format of current_stack entries.
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 recommends what to add next given current stack and new context, with an example. It distinguishes from sibling tools (explain_why, recommend_for_goal, recommend_for_project) by focusing on 'next additions' in a mid-project advisor role.
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 as a 'mid-project advisor' and provides an example, giving clear context. However, it does not explicitly state when not to use this tool or contrast with alternatives, slightly limiting guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
2 tool updates
v0.3.0- Added
recommend_for_goal - Changed
recommend_next3 fields changed- added
Input schema / properties / current_stack / itemsAdded value: +{ + "type": "string" +} - changed
Input schema / properties / current_stack / typePrevious value: -"string"New value: +"array" - added
Input schema / properties / session_fileAdded value: +{ + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "default": null +}
3 tool updates
v0.1.2- First observed
explain_why - First observed
recommend_for_project - First observed
recommend_next
TDQS
Scored across 4 tools
Each tool has a distinct purpose: explaining a specific server, recommending for a multi-part goal, for a full project, or for next steps. No overlap.
All names follow verb_noun pattern (explain_why, recommend_for_goal, etc.), but 'explain_why' uses 'why' as a noun, slightly deviating from standard action-object convention.
Four tools cover the essential recommendation scenarios without redundancy. The count is well-scoped for the server's purpose.
The set covers main use cases: explaining, recommending for goals, projects, and next steps. Missing features like comparing or updating recommendations, but not critical.
Maintenance
Related MCP Connectors
The MCP server index that vets servers, not just lists them. Advisory screen before you install.
MCP server for developer documentation, generated by doc2mcp.
Search a directory of 800+ open-source MCP servers: find, compare and get install configs.
MCP registry & directory: search, find & install 31k+ MCP servers & tools. Catalog and marketplace.