Intel Overdrive

Real-time intelligence about new MCP servers, AI coding tools, and ecosystem updates from 1,000+ monitored sources — delivered directly into Claude Code

GitHub

Intel Overdrive

npm version npm downloads GitHub stars License

Your AI coding agent doesn't know what shipped last week. Intel Overdrive gives it a live feed of breaking changes, new tools, and security alerts from 1,100+ sources — so it stops generating wrong code.

Get started

Paste into Claude Code:

npx intel-overdrive setup

No email, no account, no restart. Works immediately.

Then ask your agent anything about the AI ecosystem:

You: "What did I miss this week in AI coding?"

Agent runs intel-overdrive feed --days 7:

  Claude Code now supports scheduled tasks and multi-agent code review
  Gemini 2.5 Flash — new "thinking budget" for fine-grained reasoning control
  MCP slash-commands workflow migration — update branch protection rules
  OpenClaw 2026.3.24 — breaking changes to Gateway/OpenAI compatibility
  How to split work between Claude Code and Codex in real projects

No more scrolling Twitter, Reddit, and newsletters. Your agent already knows.

[!TIP] Also available via skills.sh: npx skills add Looney-tic/agent-skills --skill intel-overdrive -g -y

What you can ask

Instead of...Ask your agent
Scrolling Twitter for AI news"What's new in AI coding this week?"
Checking changelogs before upgrading"Any breaking changes I should know about?"
Googling "best tool for X""What's the best MCP server for databases?"
Reading newsletters you're behind on"Catch me up on Claude Code — what changed?"
Wondering if your SDK is outdated"Is the Anthropic SDK I'm using still current?"

Your agent calls it automatically when you ask about AI tools, SDKs, or frameworks. Or use the CLI directly:

intel-overdrive feed --days 7           # what's new this week
intel-overdrive breaking                # breaking changes to watch for
intel-overdrive search "Claude Code hooks best practices"

How it works

  1. Skill tells your agent when to query (installed to ~/.claude/skills/)
  2. CLI does the querying via Bash — fast, authenticated, no background process
  3. 1,100+ sources monitored: 22k+ GitHub repos, 280+ RSS feeds, npm, PyPI, Reddit, HN, arXiv

No MCP server required. No restart. The agent runs intel-overdrive search "query" whenever the skill triggers.

[!NOTE] Want the structured MCP tool in your tool list? Run intel-overdrive mcp-enable — optional, requires a Claude Code restart.

Why not web search?

Agent web searchIntel Overdrive
Speed10-30s of Googling and scrapingOne call, instant
CostMultiple tool calls, burns tokensSingle call, pre-compressed
ReliabilityScrapes fail, results outdatedPre-indexed, quality-scored
QualityCan't rank a 30k-star SDK from a weekend projectStars, maturity labels, significance tiers

Coverage

1,100+ sources polled every 15 minutes. 49,000+ items classified and searchable.

Sources include GitHub repos (22k+), RSS/Atom feeds (280+), vendor MCP servers (30+), Reddit, Hacker News, Bluesky, npm, PyPI, arXiv, VS Code Marketplace, and MCP registries. Every item is auto-classified by type and significance level.

CLI reference

CommandDescription
intel-overdrive setupRegister API key, install CLI, add skill
intel-overdrive search "query"Search for tools, docs, best practices
intel-overdrive feed [--days N]Recent updates sorted by significance
intel-overdrive breaking [--days N]Breaking changes and deprecations
intel-overdrive mcp-enableOptional: register as MCP server
intel-overdrive --versionShow version

Self-host

git clone https://github.com/Looney-tic/intel-overdrive.git && cd intel-overdrive
docker compose up -d && cp .env.example .env && alembic upgrade head

Requires Python 3.12+, PostgreSQL with pgvector, Redis, Voyage AI key, Anthropic key. API docs →

Why I built this

I got frustrated. Every morning I was scrolling Twitter, checking Reddit, skimming newsletters — just to stay current on what's new in the AI coding ecosystem. It felt like a part-time job.

Then I realized: when I asked Claude Code about recent developments, it just started Googling for me. Scraping random pages, burning tokens, returning outdated results. My AI coding agent couldn't tell me what shipped last week.

So I built Intel Overdrive. 1,100+ sources, AI-classified, quality-scored, queryable in a single call. Now when I ask "what's the best MCP server for databases?" — my agent already knows.

— Tijmen

License

Elastic License 2.0 — free to use, modify, and self-host.

Related Servers