KORE — Claude Code MCP server: what it does, how to install
The KORE MCP server for Claude Code provides 14 agent tools for model routing, context compression, code isolation, and hallucination checks.
no license 832 installs updated 2026-06-30
claude mcp add --transport http kore-api https://kore-api--amonmaly-33.run.toolsWhat KORE does
This server exposes an orchestration layer designed for AI agents. The route tool scans an 11-provider fallback sequence to dispatch tasks to the lowest-cost LLM that satisfies quality constraints. Context size shrinks up to five times via compress, while guard and verify inspect model output against source files to detect ungrounded claims.
Execution safety relies on normalize to filter prompt injections and bad characters. Code runs inside sandbox with a 30-second cap and automatic file cleanup. Docker is unnecessary here. Agent memory stores vectors in Qdrant via all-MiniLM embeddings, and tools like provenance generate HMAC-SHA256 audit records aligned with the EU AI Act.
An API key is required and generated instantly via POST /v1/register. New accounts receive 500 free credits with no credit card attached. A standard six-step processing chain consumes 7 credits per execution.
What it helps with
- Cut context token overhead using semantic text compression
- Dispatch inference requests to the cheapest capable model provider
- Flag hallucinations by cross-checking model drafts against source files
- Execute untrusted Python scripts safely without configuring Docker
- Produce tamper-evident HMAC-SHA256 logs for regulatory compliance
Example prompts for the agent
- Check the generated summary against
sources.txtwithverifytool - Compress this raw log text using
compressbefore passing it to the context - Run this data formatting script via
sandboxand return the stdout
How to install KORE
Platform CLI: remote server
claude mcp add --transport http kore-api https://kore-api--amonmaly-33.run.toolsPlatform CLI: remote server
codex mcp add kore --url https://kore-api--amonmaly-33.run.toolsConfig file
{
"mcpServers": {
"kore": {
"url": "https://kore-api--amonmaly-33.run.tools"
}
}
}Config file
{
"servers": {
"kore": {
"type": "http",
"url": "https://kore-api--amonmaly-33.run.tools"
}
}
}Platform CLI: remote server
gemini mcp add -t http kore https://kore-api--amonmaly-33.run.toolsStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about KORE
How do I install KORE in Claude Code?
Run claude mcp add --transport http kore-api https://kore-api--amonmaly-33.run.tools in your terminal. This configures the remote HTTP endpoint directly.
How does the credit system work for KORE?
Calling POST /v1/register gives you an instant key with 500 free credits and no payment details needed. A standard pipeline cycle costs 7 credits, while paid plans start at $9 for 10,000 credits.
Does the sandbox tool require a local Docker setup?
No local Docker daemon is needed. The server runs Python scripts in an isolated environment with a 30-second timeout and cleans up temporary files automatically.