Inquisitor MCP server
{
"mcpServers": {
"1111111111111111111114olvt2-inquisitor": {
"command": "uvx",
"args": [
"inquisitor-mcp"
],
"env": {
"BRAVE_API_KEY": "<brave_api_key>",
"SEARXNG_URL": "https://..."
}
}
}
}{
"mcpServers": {
"1111111111111111111114olvt2-inquisitor": {
"command": "uvx",
"args": [
"inquisitor-mcp"
],
"env": {
"BRAVE_API_KEY": "<brave_api_key>",
"SEARXNG_URL": "https://..."
}
}
}
}{
"mcpServers": {
"1111111111111111111114olvt2-inquisitor": {
"command": "uvx",
"args": [
"inquisitor-mcp"
],
"env": {
"BRAVE_API_KEY": "<brave_api_key>",
"SEARXNG_URL": "https://..."
}
}
}
}{
"servers": {
"1111111111111111111114olvt2-inquisitor": {
"command": "uvx",
"args": [
"inquisitor-mcp"
],
"env": {
"BRAVE_API_KEY": "${input:brave_api_key}",
"SEARXNG_URL": "${input:searxng_url}"
}
}
},
"inputs": [
{
"id": "brave_api_key",
"type": "promptString",
"description": "Brave api key",
"password": true
},
{
"id": "searxng_url",
"type": "promptString",
"description": "Searxng url",
"password": true
}
]
}{
"mcpServers": {
"1111111111111111111114olvt2-inquisitor": {
"command": "uvx",
"args": [
"inquisitor-mcp"
],
"env": {
"BRAVE_API_KEY": "<brave_api_key>",
"SEARXNG_URL": "https://..."
}
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall uvx inquisitor-mcpversion 0.2.2
What it does
Inquisitor is an MCP server that implements hypothesis-driven problem solving for AI agents. It provides tools for web search, codebase analysis, code tracing, and project scaffolding, combined with a methodology that probes for information, eliminates false paths, and escalates investigation depth only when evidence requires it.
- Investigate code failures by forming hypotheses, running cheap probes to confirm or eliminate them, and escalating depth only when initial evidence is insufficient.
- Search codebases and web resources selectively when local analysis cannot resolve a question, rather than treating external search as a default action.
- Track investigation state across multiple interactions, recording which hypotheses have been falsified and why, to avoid repeating failed approaches.
Tools this server exposes
| Tool | Does |
|---|---|
| probe | Perform the cheapest action that could confirm or eliminate the leading hypothesis. |
| web_search | Search the web when local evidence is insufficient to resolve the current question. |
| codebase_scan | Analyze a project's structure and contents to understand its architecture. |
| code_trace | Trace execution paths through code to identify where behavior occurs. |
| project_scaffold | Generate or initialize project structure and configuration. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.