Tools — 46 posts · Page 3
Is Gemini CLI still usable? Free ended, API keys work
Since June 18, free and personal AI Pro/Ultra accounts can't use Gemini CLI and must move to Antigravity CLI. Paid keys and company licenses are unaffected.
Fence in your AI agent first with NVIDIA OpenShell
OpenShell turns a sandboxed agent's blocked connections into policy: denial log, draft rule, risk check, human approval. Covers defaults and auto-approve gaps.
Terminal agent oh-my-pi adds IDE tools and a watcher model
We re-counted oh-my-pi's edit-format claim against its 16-model table, then checked LSP, subagents, the advisor model and the default yolo approval mode.
nanobot, a small-core AI agent for chat and schedules
nanobot, an MIT open-source agent, plugs chat apps, tools, memory and schedules into one gateway. We take it apart and cover the two locks that start off.
13 free Rust apps that rebuild Photoshop and Premiere
The 13 Crafting Apps mapped by the work they replace, self-measured readiness (40-62%) and stars, plus what to try on a copy and which folder to give an agent.
Cherry Studio runs OpenAI, Anthropic and local models
Cherry Studio mapped by data direction: entrance (providers, imports), inside (chat, knowledge bases, agents) and exit (API gateway, channels, backups).
This MCP server maps your code so agents read fewer files
codebase-memory-mcp's 10x token saving came with quality falling from 0.92 to 0.83 in a 31-language test. How 120x became 10x, plus setup and re-measuring.
Download open models or try them on ModelScope's free API
Budget ModelScope's free API from 250 cubes, 2,000 calls per account and 200 per model, then compare job bills, provider routing and weight download costs.
HeyGen's open-source HyperFrames turns HTML into video
Build a 6-second title scene in one HTML file with HeyGen's open-source HyperFrames, then follow the frame math, lint codes, render time and render paths.
Run local models behind Claude Code with Unsloth Desktop
From installing Unsloth Desktop to the first unsloth start claude run: fixes for edits that don't land and for slowdowns, sandbox levels, and open-port risks.