MCP server bridging OpenGrok search engine with AI for instant context across massive codebases
MCP server bridging OpenGrok search engine with AI for instant context across massive codebases
Install OpenGrok MCP from the VS Code Marketplace, or search "OpenGrok" in the Extensions panel.
Compare MCP servers by category, or browse hosted MCP servers if you prefer an endpoint over a local process.
Check the linked publisher documentation and your connected client's tool list. Names and availability can vary with the server version and account permissions.
| Tool | What it does |
|---|---|
| opengrok_get_symbol_context | 1) searches definition, 2) reads source, 3) fetches headers, 4) gets references |
| opengrok_search_and_read | 1) executes search, 2) immediately fetches surrounding code context |
| opengrok_batch_search | Combines 2-5 individual search queries; deduplicates file:line hits across queries |
| opengrok_index_health | Checks latency, backend connectivity, staleness score, and latency trend |
| opengrok_what_changed | Recent line changes grouped by commit — author, date, SHA, changed lines with context. Parameters: project, path, since_days |
| opengrok_dependency_map | BFS traversal of #include/import chains up to configurable depth (1–3); directed graph with uses/used_by |
| opengrok_search_pattern | Regex code search via regexp=true; returns file:line:content matches |
| opengrok_blame | Git blame with line range (start_line/end_line); returns author, date, commit per line *(v5.6+)* |
| opengrok_call_graph | Call chain tracing via OpenGrok API v2 /symbol/{name}/callgraph (requires OPENGROK_API_VERSION=v2) |
| opengrok_get_file_diff | Unified diff between two revisions with full context lines — shows surrounding code so AI understands *why* a change was made; use opengrok_get_file_history to discover revision hashes |
| opengrok_memory_status | Shows both memory files (status, bytes, 3-line preview) — helps LLM decide whether to read |
| opengrok_read_memory | Read active-task.md or investigation-log.md from the Living Document memory bank |
| opengrok_update_memory | Write or append to memory files; auto-timestamps investigation-log.md entries |
{
"mcpServers": {
"opengrok": {
"command": "npx",
"args": ["-y", "opengrok-mcp-server"],
"env": {
"OPENGROK_HTTP_PORT": "your-value",
"OPENGROK_BASE_URL": "your-value",
"OPENGROK_HTTP_AUTH_TOKEN": "your-value",
"OPENGROK_HTTP_CLIENT_ID": "your-value",
"OPENGROK_HTTP_CLIENT_SECRET": "your-value"
}
}
}
}Add to claude_desktop_config.json, then restart Claude Desktop.
| Variable | Description | Required |
|---|---|---|
| OPENGROK_HTTP_PORT | Configuration value read at startup. | Optional |
| OPENGROK_BASE_URL | Endpoint or connection string the server talks to. | Yes |
| OPENGROK_HTTP_AUTH_TOKEN | Credential the server authenticates with. | Yes |
| OPENGROK_HTTP_CLIENT_ID | Configuration value read at startup. | Optional |
| OPENGROK_HTTP_CLIENT_SECRET | Credential the server authenticates with. | Yes |
Retrieve public web pages as readable Markdown with the reference Fetch MCP server, including pagination for longer responses.
Use Firecrawl MCP to retrieve website content and extract information for research through Firecrawl’s API-backed web tools.
MCP (Model Context Protocol) server for FineData web scraping API.
An MCP server for privacy-focused web search and article scraping using SearXNG instances.
AI visibility checker: GPTBot vs OAI-SearchBot, robots.txt audit, raw-HTML schema. 165 crawlers.
Real-time web intelligence for AI agents. 11 tools, no API keys. GitHub, HN, Reddit, arXiv & more.