MCP server for multi-Google-account access (Gmail, Calendar, Drive) [BETA - Google OAuth approval pending, by-invite access]
Ask AI about Google
Powered by Claude Β· Grounded in docs
I know everything about Google. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Reviews
Documentation
antidrift
Company brain for you and your AI agents. Shared knowledge, shared skills, no more passing files around.
Works with Claude Code, Claude Cowork / Desktop, OpenAI Codex, Cursor, and Google Antigravity.
The Problem
Every time you start a new session, your AI agent has no idea who you are, what your company does, who your customers are, or how you work. You repeat yourself constantly. Context drifts. Knowledge gets lost.
The Fix
Antidrift is a git repo that gives your AI agents full context on your company. Product, customers, stack, positioning, decisions, standards β all of it. Clone it on any machine and your agent knows everything.
Two things in the box:
- The brain β markdown files organized by department. Your agent reads them automatically.
- Skills β slash commands that do real work:
/write,/review,/icp,/prep,/tps, and more.
Install
npm install -g @antidrift/cli
Or via Python:
pip install antidrift
Works on macOS, Linux, and Windows (CMD, PowerShell, Git Bash).
Quick Start
antidrift init
That's it. Your agent loads the brain and knows where everything is. Type /ingest <path> to start building context from your existing files.
What's In the Brain
your-brain/
βββ CLAUDE.md # Company overview, priorities, navigation
βββ AGENTS.md # Same content β for Codex users
βββ product/ # What you're building, roadmap
βββ customers/ # Account data, meeting notes, agreements
βββ marketing/ # Positioning, brand voice, campaigns, ICPs
βββ sales/ # Pipeline, prospects, outreach
βββ engineering/ # Architecture, coding standards, specs
βββ finance/ # Revenue, expenses
βββ decisions/ # Why we made the choices we made
βββ projects.md # Project registry with repo URLs
Add a department by creating a directory. Nest as deep as you want. Your agent discovers it all.
Skills
Skills are slash commands. Type / to see them.
Core (ships with every brain)
| Skill | What it does |
|---|---|
/ingest <path> | Import files and directories into the brain |
/push | Commit and push changes (syncs brain files, merges additively) |
/refresh | Pull latest changes from remote |
/onboard | Walk a new person through everything |
/publish <skill> | Share a skill with the community |
Community Skill Packs
antidrift skills add essentials # decision, recap, write, prep, followup, status, search
antidrift skills add engineering # review, spec, changelog, standards, tps
antidrift skills add customer-research # icp, voc, twins
antidrift skills add email # inbox (classify, move, archive, clean)
antidrift skills add legal # legal document generator
antidrift skills add all # everything
| Pack | Skills |
|---|---|
| essentials | /decision β log decisions with context and reasoning |
/recap β what changed since last time | |
/write β write anything using brand voice + brain context | |
/prep β prep for a customer call from account data | |
/followup β draft follow-up emails after meetings | |
/status β brain health, freshness, gaps | |
/search β search across the entire brain | |
| engineering | /review β code review using team standards from the brain |
/spec β technical spec using architecture context | |
/changelog β generate changelog from git history | |
/standards β generate coding standards from existing code | |
/tps β TPS report from git history (yes, that TPS report) | |
| customer-research | /icp β build research-backed Ideal Customer Profiles |
/voc β Voice of Customer analysis in their actual language | |
/twins β create AI twin personas for market testing | |
/inbox β classify, label, move, archive, and clean your inbox (requires mcp-gmail) | |
| legal | /legal β generate NDAs, subscription agreements, PSAs from templates |
Browse what's available:
antidrift skills list
How It Works
Your Agent Reads the Brain Automatically
Every directory has a brain file (CLAUDE.md for Claude Code, AGENTS.md for Codex). Your agent loads it when working in that area. Only what's relevant gets loaded β not the whole brain at once.
Skills Are Just Markdown
A skill is a directory with a SKILL.md file:
---
name: my-skill
description: What this skill does and when to use it
---
Instructions for your agent go here. Markdown.
No build step. No framework. No config.
Two Layers of Skills
Company skills live in the brain repo at .claude/skills/. Everyone gets them via git.
Personal skills live on your machine at ~/.claude/skills/. Never shared unless you choose to. Personal skills override company skills if they share a name.
When a personal skill is useful for the team: copy it to the brain repo and /push.
Keeping the Brain Efficient
Brain files are indexes, not documents. Keep them under 250 lines β summaries and pointers.
Heavy reference material goes in _reference/ subdirectories. Your agent won't read it unless asked.
marketing/
βββ CLAUDE.md # Summary β auto-loaded (~150 lines)
βββ pricing-structure.md # Active reference
βββ _reference/ # Heavy files β only on demand
βββ twin-tests/
βββ voc/
Supported Platforms
Antidrift works across all major AI coding platforms on the same brain.
| Platform | Brain File | Config File |
|---|---|---|
| Claude Code | CLAUDE.md | .mcp.json |
| Claude Cowork / Desktop | CLAUDE.md | claude_desktop_config.json |
| OpenAI Codex | AGENTS.md | β |
| Cursor | AGENTS.md | β |
| Google Antigravity | GEMINI.md | β |
antidrift initcreates CLAUDE.md, AGENTS.md, and GEMINI.md/pushsyncs them before every commit/ingestcreates all brain files for every department- Community skills compile to the right format on install
- Teams can mix platforms β everything stays in sync
Skill Compiler
Convert skills between platforms:
antidrift cross-compile .claude/skills/my-skill --to codex
antidrift cross-compile .agents/skills/my-skill --to claude
Connect Services
Connect external services as MCP servers β your AI agent gets live access to your tools.
antidrift connect google # Google Workspace (all-in-one)
antidrift connect gmail # Gmail only
antidrift connect drive # Drive, Docs, Sheets
antidrift connect calendar # Calendar only
antidrift connect stripe # Stripe
antidrift connect attio # Attio CRM
antidrift connect github # GitHub
Platform targeting
By default, connect installs for Claude Code. Use flags to target other platforms:
antidrift connect stripe --cowork # Claude Cowork / Desktop
antidrift connect stripe --all # All detected platforms
antidrift connect stripe --claude-code # Claude Code (explicit)
Packages
| Package | What It Does |
|---|---|
@antidrift/cli | Unified CLI |
@antidrift/core | Brain + core skills + setup |
@antidrift/skills | Community skill registry |
@antidrift/mcp-google | Google Workspace β all-in-one (Sheets, Docs, Drive, Gmail, Calendar) |
@antidrift/mcp-gmail | Gmail (search, read, send, reply, labels, archive, drafts) |
@antidrift/mcp-drive | Drive + Docs + Sheets (files, folders, docs, spreadsheets) |
@antidrift/mcp-calendar | Calendar (upcoming, search, create, today) |
@antidrift/mcp-stripe | Stripe (invoices, customers, subscriptions, charges, payment links) |
@antidrift/mcp-attio | Attio CRM (people, companies, deals, tasks, notes) |
@antidrift/mcp-github | GitHub (repos, issues, PRs, actions, releases, traffic) |
Learn More
Contributing
See CONTRIBUTING.md for guidelines on sharing skills and developing core.
License
MIT
