himalaya-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@himalaya-mcplist my unread emails in the inbox"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
himalaya-mcp
The original Himalaya CLI as automatically generated MCP tools. Upstream defines commands, arguments, aliases, and Help. This adapter serializes structured inputs and runs the unmodified binary; it does not implement another email client.
Use Node 22.19 or newer; Node 24 LTS is recommended. An installed npm package needs no Rust, Python, native parser helper, or compilation. First serve downloads the matching official binary to a private cache and verifies both archive and executable SHA-256. A compatible local binary can be selected explicitly.
Install and connect
Confirm that the selected version exists on npm before connecting. If it is unavailable, use a verified package artifact from the release workflow; the configuration below is an example, not evidence of publication.
After publication, start with one command (configure accounts in Himalaya first):
npx -y himalaya-mcp@latest
npx -y himalaya-mcp@latest doctor --jsonPin an available release in your MCP client:
{
"mcpServers": {
"email": {
"command": "npx",
"args": [
"--yes",
"himalaya-mcp@2.2.1-adapter.0.1.4",
"serve",
"--config",
"/absolute/path/to/config.toml",
"--operation-dir",
"/absolute/private/path/to/operations"
]
}
}
}Replace the version with an available release and the path with your existing native Himalaya configuration. The adapter does not register OAuth applications, obtain credentials, or configure accounts. Account and backend selection remain native CLI parameters.
Once installed:
himalaya-mcp describe --json
himalaya-mcp doctor --json --config /absolute/path/to/config.toml
himalaya-mcp serve --config /absolute/path/to/config.tomldescribe lists generated tools and Help without executing Himalaya. doctor inspects local binary metadata, catalog integrity, and explicit configuration/policy paths; it neither downloads a binary nor connects to an email account. It can report binary_missing before first serve.
For a local HTTP client:
himalaya-mcp serve --transport http --host 127.0.0.1 --port 3000 --config /absolute/path/to/config.tomlThe endpoint is http://127.0.0.1:3000/mcp. HTTP has Host/Origin checks, no built-in user authentication. Use only trusted clients and let your private deployment or tunnel enforce access. One process represents one trusted user's configured accounts.
Related MCP server: Systems Manager
Calling tools
The MCP server provides common file and send-verification instructions during initialization, separate from automatically exported native Help. Clients should apply those instructions together with each tool schema.
Every native call requires a unique, stable request_id. The server saves an operation receipt before execution; a slow call returns its ID after approximately two seconds while execution continues. After a lost response, query himalaya_mcp_operation_status with the original request_id or returned operation.id; prefer include_result=false for completion checks without large output. Use himalaya_mcp_operations_list if neither identifier is available. Identical requests reuse the retained receipt rather than executing again; changed inputs under the same ID are rejected. Deduplication is bounded by history retention (up to 24 hours / 128 records). Missing history, interrupted results, or unknown never justify automatic resending. Native exit 0 confirms completion or backend acceptance for sending, not recipient delivery.
--operation-dir selects private persistent metadata; raw inputs, URLs, and email content are not saved. Full output is temporary in-memory data. Use one directory per live server, including separate diagnostic instances. A live PID owner blocks concurrent startup; recovered incomplete operations remain unknown without replay. See the operation and file contract.
Tools accept structured native params plus generated top-level file fields such as attach. Declared file arguments are not duplicated under params. A file field takes client objects containing file_id and download_url, identified through openai/fileParams metadata. The generic runtime imports each complete file and binds it to the original CLI. Shared byte uploads, stdin mail inputs, and inline raw-mail inputs have been removed from the public API.
A compatible client's file forwarding is required; a file ID alone, client path, or resource URI is not an input. If that channel is missing, report the client limitation without reconstructing files or trying workspace transfers. Complete raw mail is imported as one .eml object. Operation responses preserve available native exit status and stdout/stderr, and output resources allow retrieval without promising automatic re-input. See the file contract for schemas, limits, and client verification.
An optional, personally maintained YAML policy excludes operations you choose. It is never installed automatically; sending and attachments are not excluded by the supplied example. Runtime tool availability filtering hides wholly excluded and terminal-only commands; conditional exclusions are checked on each call. Read policy behavior before enabling --policy. Annotations describe effects and do not prove approval.
The generated schema and server-side bridge do not establish that a live ChatGPT connection forwards files correctly; real client acceptance must be verified separately.
The release manifest targets macOS arm64/x64, Linux arm64/x64, and Windows x64. Actual platform results belong to CI and release artifacts; a target listed here is not a claim that it has been tested successfully.
Develop and maintain
See CONTRIBUTING.md, architecture, and the release and recovery procedure. Development requires Rust to reflect and test upstream Clap definitions; end-user packages do not.
This server cannot be deployed
Maintenance
Related MCP Connectors
Your agent needs a mailbox of its own — to receive, thread, draft and send, with attachments, without borrowing your personal inbox or your company's SMTP. **What you can ask for** • "Create an inbox for this agent and tell me its address." • "Read the new messages in this thread and draft a reply." • "Send this message with the attachment and wait for the response." • "Search this inbox for everything from that domain." • "Show delivery metrics and the events on this inbox." **How to use it** Point any MCP client at https://mcp.aisa.one/mail/mcp and sign in with OAuth — there is no key to create or paste. 49 tools: create and delete inboxes, list and read messages, raw message bodies, attachments, threads, drafts and draft attachments, send and reply, message search, inbox events, metrics, and list entries — reads and writes. **Why this rather than the source** A real inbox an agent owns, rather than an SMTP credential it borrows from a human. **It is also a door to the rest** The same login reaches 26 sources and 580+ operations. Find the contact elsewhere in the catalogue, then write to them from here — without adding a second server. **What it costs** Finding and inspecting an operation is free. Running one is billed per call at API prices, with no seat and no monthly minimum, and every call takes max_price_usd so an agent cannot overspend by accident. **Where else it reaches** https://mcp.aisa.one/sales/mcp finds the person to write to.
Remote MCP for Android CLI agent build gate, structured receipts, audit logs, and reviewer-ready evi
Email infrastructure for AI agents — send, receive, search, and reply to email over MCP.
Publish HTML, files, or a URL to a permanent public URL, then update it — from any MCP agent.
Related MCP Servers
- AlicenseNot gradedqualityAmaintenanceEnables users to define and run MCP tools using declarative YAML configs with built-in trust enforcement, credential brokering, and tamper-evident audit logging.14MIT
- AlicenseNot gradedqualityAmaintenanceEnables AI agents to manage system updates, application installations, and remote host orchestration through MCP tools.3MIT
- AlicenseNot gradedqualityAmaintenanceCLI + MCP toolkit for Anypoint Platform that enables deploying applications, tailing logs, pulling metrics, and managing API specs with production safety nets.33 npm1MIT
- AlicenseNot gradedqualityBmaintenanceProvides OAuth-based MCP integration and client-side tooling for connecting agents to a remote service, with CLI and plugin support.MIT