When tool count exceeds LLM context limits, accuracy collapses (248 tools → 12%). graph-tool-call builds a tool graph from OpenAPI/MCP specs and retrieves multi-step workflows via hybrid search (BM25 + graph traversal + embedding), recovering accuracy to 82% with 79% fewer tokens. Zero dependencies. Also works as an MCP Proxy — aggregate multiple MCP servers behind 3 meta-tools.
Run the underlying MCP server directly — without Agent-CoreX routing.
npx -y graph-tool-callAgent-CoreX intelligently routes which tools from SonAIengine/graph-tool-call your AI agent actually needs per request — reducing token usage and cost.
agent-corex.json
{
"mcpServers": {
"graph-tool-call": {
"command": "npx",
"args": [
"-y",
"graph-tool-call"
]
}
}
}Connect in 2 minutes. Only pay for the tools your agent actually uses.
A collection of reference implementations for the Model Context Protocol (MCP), showcasing its features and SDK usage.
A CLI inspector for the Model Context Protocol that allows running MCP servers and interacting with various tools and resources.
A project related to the Model Context Protocol, potentially involving data processing or context management.
Recommended: intelligent tool routing, lower costs
uvx agent-corex mcp add graph-tool-callAgent-CoreX selects only the tools your agent needs, cutting token usage by up to 60%.
Use SonAIengine/graph-tool-call directly without routing
npx -y graph-tool-callNew to Agent-CoreX? View setup guide →