Skip to content

Instantly share code, notes, and snippets.

@Evan-Kim2028
Evan-Kim2028 / grok-mcp-setup.md
Created June 5, 2026 17:03
Grok CLI as MCP Server for Claude Code (X search, no API key)

Grok CLI as MCP Server for Claude Code (No API Key)

Use your existing Grok Build OAuth session to add X search and Grok as an MCP tool inside Claude Code — no API key required.

Prerequisites

  • Grok Build CLI installed and authenticated via grok login
  • Claude Code installed
  • Python 3.10+ with uv available (pip install uv or brew install uv)
@Evan-Kim2028
Evan-Kim2028 / takopi-composer-pi-gist.md
Created May 23, 2026 16:05
Composer 2.5 + pi inside Takopi (Telegram) — setup guide

Composer 2.5 + pi inside Takopi (Telegram)

How to run Cursor Composer 2.5 as the Takopi engine via pi + pi-cursor-sdk, without the Takopi control plane.

Tested on Linux with Takopi installed at /opt/takopi-runtime, pi via nvm, and Telegram transport.


Architecture (what talks to what)

@Evan-Kim2028
Evan-Kim2028 / kimi-k2.6-droid-setup.md
Created May 19, 2026 17:44
Using a Kimi Code subscription (Kimi K2.6) with Factory's Droid CLI via BYOK + local User-Agent proxy

Using a Kimi Code subscription (Kimi K2.6) with Factory's Droid CLI (BYOK, no Droid subscription)

This guide sets up Kimi K2.6 as a custom BYOK model in the Droid CLI using a Kimi Code subscription key (the coding-plan key, sk-kimi-...), so you don't need a paid Droid subscription. Inference bills to your Kimi Code subscription.

Works the same whether you run droid directly or inside a terminal multiplexer like herdr (herdr just hosts the droid process in a pane — no special integration needed).

@Evan-Kim2028
Evan-Kim2028 / gist:818ec5bc101b44c426c0c2ec22af9ff1
Created April 7, 2026 14:39
Local push-to-talk dictation on Linux (System76 / Pop!_OS) using faster-whisper + GPU
# Local Push-to-Talk Dictation on Linux (System76 / Pop!_OS)
Uses [faster-whisper](https://github.com/SYSTRAN/faster-whisper) + your NVIDIA GPU for local, private speech-to-text.
Hold **Right Alt** to record, release to transcribe and type into any focused window.
Tested on: System76 laptop, Pop!_OS 22.04, NVIDIA RTX 4050, GNOME on X11.
---
## Prerequisites
txs_calldata: pl.LazyFrame = (
pl.scan_parquet("data/raw/transactions/*.parquet")
.select(
"block_number",
"transaction_hash",
"input",
)
.with_columns([pl.col("input").str.len_bytes().alias("calldata_size")])
.with_columns(pl.col("calldata_size") - 1)
.select("transaction_hash", "block_number", "calldata_size")
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [