| Name | consult7 JSON |
| Version |
2.1.0
JSON |
| download |
| home_page | None |
| Summary | MCP server for consulting large context window models to analyze extensive file collections |
| upload_time | 2025-09-01 21:19:12 |
| maintainer | None |
| docs_url | None |
| author | None |
| requires_python | >=3.11 |
| license | MIT |
| keywords |
code-analysis
large-context
llm
mcp
|
| VCS |
 |
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
# Consult7 MCP Server
**Consult7** is a Model Context Protocol (MCP) server that enables AI agents to consult large context window models for analyzing extensive file collections - entire codebases, document repositories, or mixed content that exceed the current agent's context limits. Supports providers *Openrouter*, *OpenAI*, and *Google*.
## Why Consult7?
When working with AI agents that have limited context windows (like Claude with 200K tokens), **Consult7** allows them to leverage models with massive context windows to analyze large codebases or document collections that would otherwise be impossible to process in a single query.
> "For Claude Code users, Consult7 is a game changer."
## How it works
**Consult7** collects files from the specific paths you provide (with optional wildcards in filenames), assembles them into a single context, and sends them to a large context window model along with your query. The result is directly fed back to the agent you are working with.
## Example Use Cases
### Summarize an entire codebase
* **Files:** `["/Users/john/project/src/*.py", "/Users/john/project/lib/*.py"]`
* **Query:** "Summarize the architecture and main components of this Python project"
* **Model:** `"gemini-2.5-flash"`
### Find specific method definitions
* **Files:** `["/Users/john/backend/src/*.py", "/Users/john/backend/auth/*.js"]`
* **Query:** "Find the implementation of the authenticate_user method and explain how it handles password verification"
* **Model:** `"gemini-2.5-pro"`
### Analyze test coverage
* **Files:** `["/Users/john/project/tests/*_test.py", "/Users/john/project/src/*.py"]`
* **Query:** "List all the test files and identify which components lack test coverage"
* **Model:** `"gemini-2.5-flash"`
### Complex analysis with thinking mode
* **Files:** `["/Users/john/webapp/src/*.py", "/Users/john/webapp/auth/*.py", "/Users/john/webapp/api/*.js"]`
* **Query:** "Analyze the authentication flow across this codebase. Think step by step about security vulnerabilities and suggest improvements"
* **Model:** `"gemini-2.5-flash|thinking"`
### Generate a report saved to file
* **Files:** `["/Users/john/project/src/*.py", "/Users/john/project/tests/*.py"]`
* **Query:** "Generate a comprehensive code review report with architecture analysis, code quality assessment, and improvement recommendations"
* **Model:** `"gemini-2.5-pro"`
* **Output File:** `"/Users/john/reports/code_review.md"`
* **Result:** Returns `"Result has been saved to /Users/john/reports/code_review.md"` instead of flooding the agent's context
## Installation
### Claude Code
Simply run:
```bash
# OpenRouter
claude mcp add -s user consult7 uvx -- consult7 openrouter your-api-key
# Google AI
claude mcp add -s user consult7 uvx -- consult7 google your-api-key
# OpenAI
claude mcp add -s user consult7 uvx -- consult7 openai your-api-key
```
### Claude Desktop
Add to your Claude Desktop configuration file:
```json
{
"mcpServers": {
"consult7": {
"type": "stdio",
"command": "uvx",
"args": ["consult7", "openrouter", "your-api-key"]
}
}
}
```
Replace `openrouter` with your provider choice (`google` or `openai`) and `your-api-key` with your actual API key.
No installation required - `uvx` automatically downloads and runs consult7 in an isolated environment.
## Command Line Options
```bash
uvx consult7 <provider> <api-key> [--test]
```
- `<provider>`: Required. Choose from `openrouter`, `google`, or `openai`
- `<api-key>`: Required. Your API key for the chosen provider
- `--test`: Optional. Test the API connection
The model is specified when calling the tool, not at startup. The server shows example models for your provider on startup.
### Model Examples
#### Google
Standard models:
- `"gemini-2.5-flash"` - Fast model
- `"gemini-2.5-flash-lite"` - Ultra fast lite model
- `"gemini-2.5-pro"` - Intelligent model
- `"gemini-2.0-flash-exp"` - Experimental model
With thinking mode (add `|thinking` suffix):
- `"gemini-2.5-flash|thinking"` - Fast with deep reasoning
- `"gemini-2.5-flash-lite|thinking"` - Ultra fast with deep reasoning
- `"gemini-2.5-pro|thinking"` - Intelligent with deep reasoning
#### OpenRouter
Standard models:
- `"google/gemini-2.5-pro"` - Intelligent, 1M context
- `"google/gemini-2.5-flash"` - Fast, 1M context
- `"google/gemini-2.5-flash-lite"` - Ultra fast, 1M context
- `"anthropic/claude-sonnet-4"` - Claude Sonnet, 200k context
- `"anthropic/claude-opus-4.1"` - Claude Opus 4.1, 200k context
- `"openai/gpt-5"` - GPT-5, 400k context
- `"openai/gpt-4.1"` - GPT-4.1, 1M+ context
With reasoning mode (add `|thinking` suffix):
- `"anthropic/claude-sonnet-4|thinking"` - Claude with 31,999 reasoning tokens
- `"anthropic/claude-opus-4.1|thinking"` - Opus 4.1 with reasoning
- `"google/gemini-2.5-flash-lite|thinking"` - Ultra fast with reasoning
- `"openai/gpt-5|thinking"` - GPT-5 with reasoning
- `"openai/gpt-4.1|thinking"` - GPT-4.1 with reasoning effort=high
#### OpenAI
Standard models (include context length):
- `"gpt-5|400k"` - GPT-5, 400k context
- `"gpt-5-mini|400k"` - GPT-5 Mini, faster
- `"gpt-5-nano|400k"` - GPT-5 Nano, ultra fast
- `"gpt-4.1-2025-04-14|1047576"` - 1M+ context, very fast
- `"gpt-4.1-nano-2025-04-14|1047576"` - 1M+ context, ultra fast
- `"o3-2025-04-16|200k"` - Advanced reasoning model
- `"o4-mini-2025-04-16|200k"` - Fast reasoning model
O-series models with |thinking marker:
- `"o1-mini|128k|thinking"` - Mini reasoning with |thinking marker
- `"o3-2025-04-16|200k|thinking"` - Advanced reasoning with |thinking marker
**Note:** For OpenAI, |thinking is only supported on o-series models and serves as an informational marker. The models use reasoning tokens automatically.
**Advanced:** You can specify custom thinking tokens with `|thinking=30000` but this is rarely needed.
## File Specification Rules
When using the consultation tool, you provide a list of file paths with these rules:
1. **All paths must be absolute** (start with `/`)
- ✅ Good: `/Users/john/project/src/*.py`
- ❌ Bad: `src/*.py` or `./src/*.py`
2. **Wildcards (`*`) only allowed in filenames**, not in directory paths
- ✅ Good: `/Users/john/project/*.py`
- ❌ Bad: `/Users/*/project/*.py` or `/Users/john/**/*.py`
3. **Must specify extension when using wildcards**
- ✅ Good: `/Users/john/project/*.py`
- ❌ Bad: `/Users/john/project/*`
4. **Mix specific files and patterns freely**
- ✅ Good: `["/path/src/*.py", "/path/README.md", "/path/tests/*_test.py"]`
5. **Common patterns:**
- All Python files in a directory: `/path/to/dir/*.py`
- Test files: `/path/to/tests/*_test.py` or `/path/to/tests/test_*.py`
- Multiple extensions: Use multiple patterns like `["/path/*.js", "/path/*.ts"]`
The tool automatically ignores: `__pycache__`, `.env`, `secrets.py`, `.DS_Store`, `.git`, `node_modules`
**Size limits:** 1MB per file, 4MB total (optimized for ~1M token context windows)
## Tool Parameters
The consultation tool accepts the following parameters:
- **files** (required): List of absolute file paths or patterns with wildcards in filenames only
- **query** (required): Your question or instruction for the LLM to process the files
- **model** (required): The LLM model to use (see Model Examples above for each provider)
- **output_file** (optional): Absolute path to save the response to a file instead of returning it
- If the file exists, it will be saved with `_updated` suffix (e.g., `report.md` → `report_updated.md`)
- When specified, returns only: `"Result has been saved to /path/to/file"`
- Useful for generating reports, documentation, or analyses without flooding the agent's context
## Testing
```bash
# Test OpenRouter
uvx consult7 openrouter sk-or-v1-... --test
# Test Google AI
uvx consult7 google AIza... --test
# Test OpenAI
uvx consult7 openai sk-proj-... --test
```
## Uninstalling
To remove consult7 from Claude Code (or before reinstalling):
```bash
claude mcp remove consult7 -s user
```
Raw data
{
"_id": null,
"home_page": null,
"name": "consult7",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.11",
"maintainer_email": null,
"keywords": "code-analysis, large-context, llm, mcp",
"author": null,
"author_email": "Stefan Szeider <stefan@szeider.net>",
"download_url": "https://files.pythonhosted.org/packages/85/ca/75b222616c9d7e9ec0fcdd30592a1bea8ee0c1793d81c20acb5ad3458697/consult7-2.1.0.tar.gz",
"platform": null,
"description": "# Consult7 MCP Server\n\n**Consult7** is a Model Context Protocol (MCP) server that enables AI agents to consult large context window models for analyzing extensive file collections - entire codebases, document repositories, or mixed content that exceed the current agent's context limits. Supports providers *Openrouter*, *OpenAI*, and *Google*.\n\n## Why Consult7?\n\nWhen working with AI agents that have limited context windows (like Claude with 200K tokens), **Consult7** allows them to leverage models with massive context windows to analyze large codebases or document collections that would otherwise be impossible to process in a single query.\n\n> \"For Claude Code users, Consult7 is a game changer.\"\n\n## How it works\n\n**Consult7** collects files from the specific paths you provide (with optional wildcards in filenames), assembles them into a single context, and sends them to a large context window model along with your query. The result is directly fed back to the agent you are working with.\n\n## Example Use Cases\n\n### Summarize an entire codebase\n\n* **Files:** `[\"/Users/john/project/src/*.py\", \"/Users/john/project/lib/*.py\"]`\n* **Query:** \"Summarize the architecture and main components of this Python project\"\n* **Model:** `\"gemini-2.5-flash\"`\n\n### Find specific method definitions\n* **Files:** `[\"/Users/john/backend/src/*.py\", \"/Users/john/backend/auth/*.js\"]`\n* **Query:** \"Find the implementation of the authenticate_user method and explain how it handles password verification\"\n* **Model:** `\"gemini-2.5-pro\"`\n\n### Analyze test coverage\n* **Files:** `[\"/Users/john/project/tests/*_test.py\", \"/Users/john/project/src/*.py\"]`\n* **Query:** \"List all the test files and identify which components lack test coverage\"\n* **Model:** `\"gemini-2.5-flash\"`\n\n### Complex analysis with thinking mode\n* **Files:** `[\"/Users/john/webapp/src/*.py\", \"/Users/john/webapp/auth/*.py\", \"/Users/john/webapp/api/*.js\"]`\n* **Query:** \"Analyze the authentication flow across this codebase. Think step by step about security vulnerabilities and suggest improvements\"\n* **Model:** `\"gemini-2.5-flash|thinking\"`\n\n### Generate a report saved to file\n* **Files:** `[\"/Users/john/project/src/*.py\", \"/Users/john/project/tests/*.py\"]`\n* **Query:** \"Generate a comprehensive code review report with architecture analysis, code quality assessment, and improvement recommendations\"\n* **Model:** `\"gemini-2.5-pro\"`\n* **Output File:** `\"/Users/john/reports/code_review.md\"`\n* **Result:** Returns `\"Result has been saved to /Users/john/reports/code_review.md\"` instead of flooding the agent's context\n\n## Installation\n\n### Claude Code\n\nSimply run:\n\n```bash\n# OpenRouter\nclaude mcp add -s user consult7 uvx -- consult7 openrouter your-api-key\n\n# Google AI\nclaude mcp add -s user consult7 uvx -- consult7 google your-api-key\n\n# OpenAI\nclaude mcp add -s user consult7 uvx -- consult7 openai your-api-key\n```\n\n### Claude Desktop\n\nAdd to your Claude Desktop configuration file:\n\n```json\n{\n \"mcpServers\": {\n \"consult7\": {\n \"type\": \"stdio\",\n \"command\": \"uvx\",\n \"args\": [\"consult7\", \"openrouter\", \"your-api-key\"]\n }\n }\n}\n```\n\nReplace `openrouter` with your provider choice (`google` or `openai`) and `your-api-key` with your actual API key.\n\nNo installation required - `uvx` automatically downloads and runs consult7 in an isolated environment.\n\n\n## Command Line Options\n\n```bash\nuvx consult7 <provider> <api-key> [--test]\n```\n\n- `<provider>`: Required. Choose from `openrouter`, `google`, or `openai`\n- `<api-key>`: Required. Your API key for the chosen provider\n- `--test`: Optional. Test the API connection\n\nThe model is specified when calling the tool, not at startup. The server shows example models for your provider on startup.\n\n### Model Examples\n\n#### Google\nStandard models:\n- `\"gemini-2.5-flash\"` - Fast model\n- `\"gemini-2.5-flash-lite\"` - Ultra fast lite model\n- `\"gemini-2.5-pro\"` - Intelligent model\n- `\"gemini-2.0-flash-exp\"` - Experimental model\n\nWith thinking mode (add `|thinking` suffix):\n- `\"gemini-2.5-flash|thinking\"` - Fast with deep reasoning\n- `\"gemini-2.5-flash-lite|thinking\"` - Ultra fast with deep reasoning\n- `\"gemini-2.5-pro|thinking\"` - Intelligent with deep reasoning\n\n#### OpenRouter\nStandard models:\n- `\"google/gemini-2.5-pro\"` - Intelligent, 1M context\n- `\"google/gemini-2.5-flash\"` - Fast, 1M context\n- `\"google/gemini-2.5-flash-lite\"` - Ultra fast, 1M context\n- `\"anthropic/claude-sonnet-4\"` - Claude Sonnet, 200k context\n- `\"anthropic/claude-opus-4.1\"` - Claude Opus 4.1, 200k context\n- `\"openai/gpt-5\"` - GPT-5, 400k context\n- `\"openai/gpt-4.1\"` - GPT-4.1, 1M+ context\n\nWith reasoning mode (add `|thinking` suffix):\n- `\"anthropic/claude-sonnet-4|thinking\"` - Claude with 31,999 reasoning tokens\n- `\"anthropic/claude-opus-4.1|thinking\"` - Opus 4.1 with reasoning\n- `\"google/gemini-2.5-flash-lite|thinking\"` - Ultra fast with reasoning\n- `\"openai/gpt-5|thinking\"` - GPT-5 with reasoning\n- `\"openai/gpt-4.1|thinking\"` - GPT-4.1 with reasoning effort=high\n\n#### OpenAI\nStandard models (include context length):\n- `\"gpt-5|400k\"` - GPT-5, 400k context\n- `\"gpt-5-mini|400k\"` - GPT-5 Mini, faster\n- `\"gpt-5-nano|400k\"` - GPT-5 Nano, ultra fast\n- `\"gpt-4.1-2025-04-14|1047576\"` - 1M+ context, very fast\n- `\"gpt-4.1-nano-2025-04-14|1047576\"` - 1M+ context, ultra fast\n- `\"o3-2025-04-16|200k\"` - Advanced reasoning model\n- `\"o4-mini-2025-04-16|200k\"` - Fast reasoning model\n\nO-series models with |thinking marker:\n- `\"o1-mini|128k|thinking\"` - Mini reasoning with |thinking marker\n- `\"o3-2025-04-16|200k|thinking\"` - Advanced reasoning with |thinking marker\n\n**Note:** For OpenAI, |thinking is only supported on o-series models and serves as an informational marker. The models use reasoning tokens automatically.\n\n**Advanced:** You can specify custom thinking tokens with `|thinking=30000` but this is rarely needed. \n\n## File Specification Rules\n\nWhen using the consultation tool, you provide a list of file paths with these rules:\n\n1. **All paths must be absolute** (start with `/`)\n - \u2705 Good: `/Users/john/project/src/*.py`\n - \u274c Bad: `src/*.py` or `./src/*.py`\n\n2. **Wildcards (`*`) only allowed in filenames**, not in directory paths\n - \u2705 Good: `/Users/john/project/*.py`\n - \u274c Bad: `/Users/*/project/*.py` or `/Users/john/**/*.py`\n\n3. **Must specify extension when using wildcards**\n - \u2705 Good: `/Users/john/project/*.py`\n - \u274c Bad: `/Users/john/project/*`\n\n4. **Mix specific files and patterns freely**\n - \u2705 Good: `[\"/path/src/*.py\", \"/path/README.md\", \"/path/tests/*_test.py\"]`\n\n5. **Common patterns:**\n - All Python files in a directory: `/path/to/dir/*.py`\n - Test files: `/path/to/tests/*_test.py` or `/path/to/tests/test_*.py`\n - Multiple extensions: Use multiple patterns like `[\"/path/*.js\", \"/path/*.ts\"]`\n\nThe tool automatically ignores: `__pycache__`, `.env`, `secrets.py`, `.DS_Store`, `.git`, `node_modules`\n\n**Size limits:** 1MB per file, 4MB total (optimized for ~1M token context windows)\n\n## Tool Parameters\n\nThe consultation tool accepts the following parameters:\n\n- **files** (required): List of absolute file paths or patterns with wildcards in filenames only\n- **query** (required): Your question or instruction for the LLM to process the files\n- **model** (required): The LLM model to use (see Model Examples above for each provider)\n- **output_file** (optional): Absolute path to save the response to a file instead of returning it\n - If the file exists, it will be saved with `_updated` suffix (e.g., `report.md` \u2192 `report_updated.md`)\n - When specified, returns only: `\"Result has been saved to /path/to/file\"`\n - Useful for generating reports, documentation, or analyses without flooding the agent's context\n\n## Testing\n\n```bash\n# Test OpenRouter\nuvx consult7 openrouter sk-or-v1-... --test\n\n# Test Google AI\nuvx consult7 google AIza... --test\n\n# Test OpenAI\nuvx consult7 openai sk-proj-... --test\n```\n\n## Uninstalling\n\nTo remove consult7 from Claude Code (or before reinstalling):\n\n```bash\nclaude mcp remove consult7 -s user\n```\n\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "MCP server for consulting large context window models to analyze extensive file collections",
"version": "2.1.0",
"project_urls": {
"Homepage": "https://github.com/szeider/consult7",
"Issues": "https://github.com/szeider/consult7/issues",
"Repository": "https://github.com/szeider/consult7"
},
"split_keywords": [
"code-analysis",
" large-context",
" llm",
" mcp"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "4e8f1555d0a626d4dc18bc967242b25cb9257acfc3135a99d51a9dd7c5cdd0f2",
"md5": "9c57e8b652f0d58a43b2ec1bdfabdbaf",
"sha256": "904c6328238d056f09a8fafb395ae2fa44d052a390e515e64066ac552505c581"
},
"downloads": -1,
"filename": "consult7-2.1.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "9c57e8b652f0d58a43b2ec1bdfabdbaf",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.11",
"size": 27813,
"upload_time": "2025-09-01T21:19:11",
"upload_time_iso_8601": "2025-09-01T21:19:11.551994Z",
"url": "https://files.pythonhosted.org/packages/4e/8f/1555d0a626d4dc18bc967242b25cb9257acfc3135a99d51a9dd7c5cdd0f2/consult7-2.1.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "85ca75b222616c9d7e9ec0fcdd30592a1bea8ee0c1793d81c20acb5ad3458697",
"md5": "93e6ba56ff709f46ca7bbbb6c1487b37",
"sha256": "c683278a9bc4575367009939483254173eab28a55e7201f93111ebc26fa3f691"
},
"downloads": -1,
"filename": "consult7-2.1.0.tar.gz",
"has_sig": false,
"md5_digest": "93e6ba56ff709f46ca7bbbb6c1487b37",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.11",
"size": 20717,
"upload_time": "2025-09-01T21:19:12",
"upload_time_iso_8601": "2025-09-01T21:19:12.679312Z",
"url": "https://files.pythonhosted.org/packages/85/ca/75b222616c9d7e9ec0fcdd30592a1bea8ee0c1793d81c20acb5ad3458697/consult7-2.1.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-09-01 21:19:12",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "szeider",
"github_project": "consult7",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "consult7"
}
