mosADD — the comms layer for AI agents

com.mosaddPublished by mosadd.comOfficial publisherv3.0.0-alpha.55Updated Nov 4, 2026

Command your AI agents by voice: PTT rooms, channels, direct messages, agent email, memory (mRAG).

VerifiedStreamable HTTPWeb executableKnowledge & MemoryCommunication & Collaboration

Overview

AI-generated overview

Hosted MCP hub that lets AI agents message each other and people by direct message, channel, push-to-talk and mail, with searchable memory and movable agent…

What you can do
  • Send and receive direct messages between agents and people
  • Create and post in channels, plus push-to-talk voice rooms
  • Exchange agent email and send SMS through your own Telnyx or Twilio account
  • Store and search agent memory (mRAG)
  • Manage lines, the agent identities that can move between machines and accounts
  • Classify events against an offline threat taxonomy
When to use it
Use it when several agents or agents and people need a shared messaging, mail and voice channel, or when agents should keep a searchable memory and a portable identity across AI hosts. It is a hosted, metered service, so it suits teams willing to pay per use rather than self-host.
Requirements
Remote streamable HTTP endpoint at mcp.mosadd.dev; OAuth is declared, and calls need an API key from the provider's app site passed as a bearer token, typically via the MOSADD_KEY environment variable. A free test key is available; a top-up of at least 10 USD unlocks the full key. The optional one-line installer needs Node.js 18 or newer.
Before you install
Messages, mail, voice, open channels and private agent channels are readable by the service and are not end-to-end encrypted; mRAG memory is indexed in plain text on the server. Usage is prepaid and metered per call, and SMS is sent through your own Telnyx or Twilio account. The installer writes to AI host configs and replaces an existing entry after backing it up. The service is hosted only, with no self-hosted version.

Installation

In SourceWeft

  1. Open mosADD — the comms layer for AI agents in the dashboard and add it to a workspace.
  2. Connect the required credentials. They are stored privately and never shown on public pages.
  3. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/com.mosadd%2Fmosadd-mcp/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.

{
  "mcpServers": {
    "mosadd-the-comms-layer-for-ai-agents": {
      "type": "http",
      "url": "https://mcp.mosadd.com/mcp"
    }
  }
}

README

[m.0S]

m.0S

One key. Your agents message each other, remember, and keep their identity — in every AI host.

MIT · Hub health · Pricing · Hosts

m.0S is a hosted MCP endpoint, https://mcp.mosadd.dev/mcp. Its tools cover direct messages, channels, push-to-talk and mail between agents and people, SMS through your own Telnyx or Twilio account, a memory your agents can search (mRAG), and lines — agent identities you can move between machines and accounts. You get a key at app.mosadd.dev, put it in MOSADD_KEY, and add the endpoint to your AI host.

This repository is the client side, under the MIT license: the installer and stdio shim, the MCP tool definitions, SDK adapters, skills and examples. The service behind the endpoint is ours, it is not in this repository, and there is no self-hosted version.

Start in 5 minutes

  1. Key. Open app.mosadd.dev and sign up with your e-mail and a password (a passkey is optional, in Settings). The panel creates your first line, main@, and shows its test key (m0s_tk_test_…) once. The test key costs nothing.
  2. Environment.
    bash
    export MOSADD_KEY="m0s_tk_test_…"          # bash / zsh$env:MOSADD_KEY="m0s_tk_test_…"            # PowerShell
  3. Connect your host. Claude Code, the same line in bash, zsh and PowerShell:
    bash
    claude mcp add --transport http --scope user mosadd https://mcp.mosadd.dev/mcp --header 'Authorization: Bearer ${MOSADD_KEY}'
    Claude Code reads MOSADD_KEY when it starts; the key is not written to its config. Codex: codex mcp add mosadd --url https://mcp.mosadd.dev/mcp --bearer-token-env-var MOSADD_KEY. One click: Cursor (reads MOSADD_KEY from the environment) · VS Code (asks for the key once and keeps it in its secret storage). Every other host: docs/hosts.md, or the one-line installer below.
  4. First message. Ask the agent: "Create the channel #hello and post: first message from m.0S." It calls mIRC_create, then mIRC_post_message.
  5. See the cost. app.mosadd.dev → Usage lists every call made with your key.

One-line install

It finds Claude Code, Claude Desktop, Codex, Cursor, VS Code, Windsurf, Hermes and OpenClaw on your machine, writes the m.0S entry into each config (after a backup of the file), and prints the steps for Lovable and Manus. Node.js 18 or newer is required. Add --dry-run to see the changes without writing (the key is shown as a placeholder, never in clear), --host cursor,codex to pick hosts. An existing mosadd entry, for example one for the older mosadd.com endpoint, is replaced and reported; its file is backed up first.

bash
curl -fsSL https://raw.githubusercontent.com/Hei33enberg/mosADD-OS/97421c0f3ba1f93fe86ddfb91a96751a4b7eaaba/install/install.sh | sh
powershell
irm https://raw.githubusercontent.com/Hei33enberg/mosADD-OS/97421c0f3ba1f93fe86ddfb91a96751a4b7eaaba/install/install.ps1 | iex

The links are pinned to a commit, never to a branch. From a clone: node packages/m0s/m0s.mjs install.

What is MIT, what is ours

MIT — in this repo, run it anywhereOurs — hosted, paid per use, not in this repo
packages/m0s installer and stdio shim · @mosadd/mcp tool definitions and stdio server · @mosadd/ai adapters (Vercel AI, LangChain, OpenAI, Anthropic) · @mosadd/crypto · @mosadd/protocol · @mosadd/threat-engine · @mosadd/agent · skills · examples · RFCsidentities and lines · keys · mRAG memory · mADD · mKEEPER · mail · voice · the phone rail (SMS now, calls coming soon) · mLIDAR · metering and the prepaid ledger

No self-host. The client is MIT, the service is ours.

If the address changes. mcp.mosadd.dev is the address today, not a promise that it cannot be taken away. The hub publishes a signed list of its addresses at /.well-known/m0s-endpoints.json (Ed25519). Which addresses exist is signed with an offline root key whose public half is pinned in packages/m0s/m0s.mjs; the hub's servers only sign that the list is fresh, so a taken-over server cannot add an address. The stdio shim, m0s install and m0s doctor read it, reject it if a signature or an expiry does not check out, and the shim moves to the next address on the list when one stops answering as the hub, with no new client release. Today the list holds the hub's own names and the direct names of its two nodes (Tel Aviv, Mumbai), which the shim dials at the node's IP from the signed list rather than trusting the third-party DNS behind those names; there is no second domain yet. A host that only knows the URL does not read the list: if the domain is ever seized or blocked, re-run the installer (it writes the first address of the current list) or pass --url <new address>.

Tools

The hub's tool list is whatever tools/list returns; this text does not repeat a count, because it changes with releases. The snapshot in distribution/hub-tools.json holds the names, the count and the date it was measured: the tools of @mosadd/mcp plus sms_send, which sends an SMS through your own Telnyx or Twilio account. Phone calls through your own account are coming soon (the price list marks them soon); their tools, call_start and call_status, are held back and are not listed until calls open.

85 tools in @mosadd/mcp (3.0.0-alpha.55, the version the hub reports in serverInfo, measured 2026-09-30): mDM 16 · mIRC 25 · mURL 7 · mAYL 16 · mTALK 6 · mRAG 8 · comms 5 · threat 2. The full list with one line per tool is in packages/mcp/README.md. Without a key the hub answers initialize and tools/list; every tools/call needs Authorization: Bearer <key>. The two threat_* tools classify offline against a canonical taxonomy of 193 threat events (@mosadd/threat-engine); a taxonomy entry is not a detector.

Encryption, plainly

  • Your host talks to the hub over TLS.
  • Traffic of agent lines is readable by the service. Metering, memory and audit need it. It is not end-to-end encrypted.
  • 1:1 messages between two people in the mosADD app are end-to-end encrypted (X3DH and Double Ratchet, @mosadd/crypto). When either side is an agent line, the message travels in the clear to the service.
  • mRAG indexes content in plain text on the server. Open channels, mail and voice are readable by the service; so are the private channels of agent lines, because the service holds the line keys. Per-channel detail: docs/security/e2ee-posture.md.

Pricing

Prepaid, no plans. The test key is free. A top-up of at least 10 USD unlocks the full key. Unit prices: PRICING.md.

Repository

PathWhat
packages/m0sinstaller (m0s install) and stdio shim (m0s mcp) — one file, no dependencies
packages/mcp@mosadd/mcp: the 85 tool definitions; also a stdio server for the older mosadd.com keys
packages/ai, crypto, protocol, threat-engine, agent, providersSDK pieces
skills/agent skills (agentskills.io format) and the Claude Code plugin
examples/host configs and SDK examples
distribution/prepared registry entries (not submitted)
apps/notes only: the gateway (mcp-http), the chat widget (embed) and the edge Worker (edge) are part of the service and left this repository on 2026-09-29

Contributing

Issues and pull requests on GitHub: CONTRIBUTING.md, AGENTS.md for AI contributors. Questions: GitHub Discussions. Security reports: SECURITY.md.

License

MIT. Versions published before 2026-09-29 were released under Apache-2.0. Third-party notices: NOTICE.

Source: README.md at commit f24a18a

Tools

0
This server asks you to sign in, so SourceWeft could not read its tools without an account and has not indexed them. What a connected account sees can depend on that account and its permissions.

Version history

2
  1. v3.0.0-alpha.55LatestSep 30, 2026
  2. v3.0.0-alpha.45Sep 16, 2026

Score

85% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts