Skip to content

MCPORTER-CLI.md

Latest commit

 

History

History
79 lines (53 loc) · 1.93 KB

MCPORTER-CLI.md

File metadata and controls

79 lines (53 loc) · 1.93 KB

Using MCPorter with Engram

MCPorter lets you call Engram's MCP tools from the command line or from TypeScript—without launching an AI client.

Prerequisites

  • Convex backend running (npx convex dev)
  • MCP server built (cd mcp-server && npm run build)
  • Environment: CONVEX_URL, ENGRAM_AGENT_ID (optional: COHERE_API_KEY)

Quick Start

# List all 52 Engram tools with TypeScript signatures
npm run mcp:list

# Or with explicit env
CONVEX_URL=https://your-deployment.convex.cloud npm run mcp:list

Call Tools

# Store a fact
npx mcporter call engram.memory_store_fact content="My preferred language is TypeScript"

# Semantic recall
npx mcporter call engram.memory_recall query="programming preferences"

# Health check
npx mcporter call engram.memory_health

With env (if not in .env or shell):

CONVEX_URL=https://your-deployment.convex.cloud ENGRAM_AGENT_ID=indy \
  npx mcporter call engram.memory_store_fact content="Remember this"

Configuration

MCPorter reads config/mcporter.json, which defines the Engram stdio server. It uses ${CONVEX_URL} and ${ENGRAM_AGENT_ID} from the environment.

To use a different config:

npx mcporter --config path/to/mcporter.json list engram

Generate Standalone CLI

Create a self-contained CLI for Engram:

npm run mcp:generate-cli
# Outputs: dist/engram-cli.js
# Run: node dist/engram-cli.js memory_store_fact content="..."

Use from TypeScript

import { createRuntime, createServerProxy } from "mcporter";

const runtime = await createRuntime();
const engram = createServerProxy(runtime, "engram");

const result = await engram.memoryStoreFact({ content: "Test fact" });
console.log(result.json());

await runtime.close();

Related