ts4k 0.1.16

Token-efficient messaging gateway for LLM agents

ts4k — Token Saver 4000

Unified messaging gateway that gives LLM agents token-efficient access to messages across Gmail, O365, and WhatsApp. Normalizes, deduplicates, filters, and formats messages for LLM context windows.

Raw HTML email: ~8,000 tokens. After ts4k: ~400 tokens. 20x reduction.

Quick Example

$ ts4k whatsnew life
1|g|[e-mail]|Q1 report draft|2026-02-24T09:15|2.1kb
2|g|[e-mail]|Re: lunch tomorrow|2026-02-24T08:30|340b
3|o|[e-mail]|Budget approval|2026-02-24T10:00|1.8kb
4|w|[phone]|Hey, running late|2026-02-24T09:45|120b

One command, three platforms, pipe-delimited output an LLM can parse in ~60 tokens. Ref numbers (1, 2, ...) let you get 1 or thread 2 without copying long IDs.

Install

pip install ts4k

Requires Python 3.12+. Creates two entry points:

• ts4k — CLI
• ts4k-mcp — MCP server for Claude Code and other MCP-compatible agents

Provider Setup

Each messaging platform needs a one-time setup. Pick the ones you use:

Provider	Guide	Auth Method	Complexity
Gmail	docs/setup-gmail.md	Google OAuth (browser)	Medium
O365	docs/setup-o365.md	Azure device code flow	Medium
WhatsApp	docs/setup-whatsapp.md	Local SQLite (no auth)	Low

You can run any combination. One provider is enough to get started.

Usage

Reading Messages

ts4k whatsnew life               # New messages since last check (keyed watermark)
ts4k updates --since 2d          # Messages from last 2 days (stateless)
ts4k updates --source g          # Gmail only
ts4k get 3                       # Read message by ref number
ts4k get g:19abc123              # Read message by full ID
ts4k thread g:thread456          # Read a thread
ts4k list -q "budget" -n 10     # Search messages
ts4k overview                    # Hierarchical cache summary
ts4k status --live               # Health + live mailbox counts

Managing Messages

Requires source level modify or higher. All actions are non-destructive and reversible.

ts4k manage archive g:abc123              # Archive a message
ts4k manage archive g:abc,g:def,g:ghi    # Batch archive
ts4k manage label g:abc --label invoices  # Add a label
ts4k manage read g:abc,g:def             # Mark as read
ts4k manage trash g:abc                   # Move to trash (recoverable)
ts4k manage list-labels g:any             # List available labels
ts4k manage archive g:abc --dry-run       # Preview without acting

Creating Drafts

Requires source level draft. ts4k never sends messages — drafts appear in your mailbox for manual review.

ts4k draft create -s g --to [e-mail] --subject "Hi" --body "Hello"
ts4k draft create -s g --reply-to g:abc123 --body "Sounds good!"  # Threaded reply

Reply drafts automatically set threading headers and blockquote the original message.

Access Levels

Each source has a permission level that controls what operations are allowed:

Level	Capabilities	OAuth Scope
readonly (default)	Read, list, search	gmail.readonly / Mail.Read
modify	+ archive, label, mark read/unread, trash	gmail.modify / Mail.ReadWrite
draft	+ create draft messages	gmail.modify / Mail.ReadWrite

Set the level when adding or updating a source:

ts4k src add g gmail email=[e-mail] level=draft

Changing the level automatically triggers re-authentication with the correct OAuth scopes.

MCP Server Mode

ts4k-mcp                              # stdio (default, for Claude Code)
ts4k-mcp --transport http --port 9000  # HTTP for other clients

10 MCP tools: updates, whatsnew, get, thread, list, overview, status, admin, manage, draft.

Architecture

Agent (Claude, etc.)
  |
  v
ts4k (normalize → filter → format)
  |── Gmail Adapter    → Google Gmail API (direct)
  |── O365 Adapter     → Microsoft Graph API (direct, httpx)
  |── WhatsApp Adapter → whatsapp-mcp (local SQLite)
  '── Future adapters  → Slack, Teams, Telegram, ...

• Adapters wrap platform APIs with a uniform interface. Each adapter supports read, manage, and draft operations gated by access level.
• Normalize strips HTML, deduplicates reply chains, collapses whitespace.
• Filter applies skip lists (senders, domains, patterns). Off by default.
• Format outputs pipe-delimited (default), JSON, or XML.

Platform failures are isolated — if one adapter is down, the others still return results.

Key Principles

• Metadata first, content on demand — default to minimum useful response
• No LLM calls inside ts4k — this is the data layer, not the intelligence layer
• ts4k never sends messages — draft creation only; the send level is defined but intentionally not implemented
• Using a command IS the side effect — watermarks update on whatsnew, no separate save step
• Format is a feature — pipe-delimited saves ~60% tokens vs JSON

Configuration

All state lives in ~/.config/ts4k/ (override with TS4K_CONFIG_DIR):

~/.config/ts4k/
  sources.json              # Configured providers (incl. access levels)
  watermarks/               # Per-key last-fetched timestamps
  contacts.json             # Cross-platform identity map
  filters.json              # Skip lists
  stats.json                # Usage counters
  cache/                    # Local message cache
  google/{email}/token.json # Gmail OAuth tokens
  microsoft/{id}/token_cache.json  # O365 MSAL tokens

Tech Stack

Python 3.12+, Google API client, MSAL, httpx, MCP SDK, html2text, beautifulsoup4.

License

MIT