Sane & Controlled AI Agent Access
Tie ALL your email accounts together in a single place—unlimited IMAP, JMAP, and Gmail accounts—and strictly control how local LLMs access your data via security policies, WORM audits, and microsecond SQLite speed.
Why Nuncio MCP is the Sovereign Standard for AI
Never grant uncontrolled direct credentials or API access to raw email providers.
1. Multi-Account Aggregation Hub
Tie ALL your email accounts—unlimited IMAP, JMAP, Gmail, Google Workspace, iCloud, and custom domain accounts—together into a single local SQLite data vault on your disk.
2. Controlled Security Gateway (Zero Direct Access)
Local LLMs and AI agents NEVER get raw passwords or direct uncontrolled access to your email accounts or Gmail servers. All access passes strictly through Nuncio's security enclave with McpAgentPolicy RBAC and automatic PII redaction.
3. WORM Audits & NSQL Controls
Every single AI agent tool call, read query, draft generation, and send action is recorded in an immutable Write Once Read Many (WORM) audit ledger sealed with HMAC-SHA256 hash chains.
4. Microsecond Local Speed
Instant local FTS5 trigram search (<10ms across 100,000+ messages) so local LLMs get sub-millisecond context retrieval without battery drain or cloud network API fees.
What is the Model Context Protocol (MCP)?
The Model Context Protocol (MCP) is an open standard created by Anthropic that allows AI applications (such as Claude Desktop, Antigravity, Cursor, or local LLM CLI agents) to safely read data and invoke actions on local services over standard JSON-RPC 2.0 stdio streams.
Ask your AI: "Find all urgent emails from Alice received today across all accounts and summarize them."
Instruct AI: "Schedule a 30-minute sync with Bob tomorrow afternoon and create the event."
Tell AI: "Create an NSQL rule to auto-archive newsletters with unsubscribe links."
Connecting Claude Desktop & Antigravity
Add the following JSON snippet to your client configuration file.
Claude Desktop / Antigravity Config (`claude_desktop_config.json`)
Executable: nuncio-mcp --stdio{
"mcpServers": {
"nuncio": {
"command": "nuncio-mcp",
"args": ["--stdio"],
"env": {
"NUNCIO_LOG_LEVEL": "info"
}
}
}
}McpAgentPolicy: Granular Security Enclave
You control exactly what data & permissions any connected AI agent receives.
Toggle fine-grained capabilities per agent: read_mail, send_mail, read_calendar, write_calendar, and manage_filters. Unauthorized operations return a 403 Forbidden error.
Before email text payloads reach the LLM context window, Nuncio's sanitize_content() engine automatically redacts Credit Cards and US Social Security Numbers ([REDACTED-SSN]).
Complete MCP Tool Matrix
Every tool available to connected AI agents over stdio streams.
List email messages with optional folder, account, and unread filters.
Dispatch outgoing emails with Human-in-the-Loop prompt confirmation.
Instant full-text FTS5 search across 100,000+ messages in under 10ms.
Export MBOX, EML ZIP, or JSON data streams programmatically.
Query immutable Write Once Read Many (WORM) audit log records.
Verify HMAC-SHA256 cryptographic hash-chain ledger integrity.
Query calendar events, schedules, and location details.
Create calendar events with start/end time conflict detection.
Inspect active server-side NSQL filter rules and execution logs.
Programmatically compile, validate, and save new NSQL filter rules.
Dry-run test NSQL filter statements against sample messages.
Inspect connected email accounts, server hostnames, and protocol status.