Back to Marketplace

Run this helper free

Answer 3 questions. Get a result in 2 minutes. Preview free.

Start free →
FREE
Verified
Grow Business

dm-other-claudes — Cross-Claude Communication

Enable Claude-to-Claude collaboration through SSH with persistent session state

Coordinating work across multiple Claude instances requires manual context passing and breaks workflow continuity.

Seamlessly delegate tasks to remote Claude instances and receive structured results without losing session context.

  • Send structured tasks to remote Claude Code instances via SSH
  • Maintain session continuity across multiple Claude interactions
  • Receive JSON-formatted responses for programmatic task chaining
  • Support non-interactive Claude execution on distributed machines

👁 3 views · 📦 0 installs

Install in one line

mfkvault install dm-other-claudes-cross-claude-communication

Requires the MFKVault CLI. Prefer MCP?

No reviews yet
🤖 Claude Code
FREE

Free to install — no account needed

Copy the command below and paste into your agent.

Instant access • No coding needed • No account needed

What you get in 5 minutes

  • Full skill code ready to install
  • Works with 1 AI agent
  • Lifetime updates included
VerifiedSecureBe the first
Ready to run

Run this helper

Answer a few questions and let this helper do the work.

Advanced: use with your AI agent

Description

# dm-other-claudes — Cross-Claude Communication Send tasks to Claude Code instances on remote machines. Enables Claude-to-Claude collaboration over SSH with session continuity. ## How It Works ``` ┌─────────────────┐ SSH + stdin ┌─────────────────┐ │ Claude (TUI) │ ──────────────────────────> │ Claude (-p) │ │ interactive │ │ non-interactive │ │ with human │ <────────────────────────── │ on remote host │ │ │ JSON response + session │ │ └─────────────────┘ └─────────────────┘ ``` 1. **Calling side** (you, in the TUI): Invokes `dm-claude.ts` via Bash tool 2. **Remote side**: Runs `claude -p --output-format json` on the target machine 3. **Session tracking**: Remote Claude's session ID is captured for follow-ups 4. **Preamble**: Auto-prepended context tells the remote Claude it's receiving a cross-Claude task ## Setup ### Prerequisites - [Bun](https://bun.sh) installed locally - [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) installed on **both** local and remote machines - SSH access to the remote machine(s) (key-based auth recommended) ### Installation > **If the user asks you to install/set up this skill**, ask them: > 1. Where are your other Claude Code agents running? (e.g., `user@hostname` or `user@ip`) > 2. Do you have SSH key-based auth to those machines, or do you need to set it up? > 3. Does the remote machine have Claude Code CLI installed and authenticated? > > Then configure the skill with their specific hosts and update the examples below accordingly. 1. Copy `dm-claude.ts` and `SKILL.md` into your Claude Code skills directory: ```bash mkdir -p ~/.claude/skills/dm-other-claudes cp dm-claude.ts SKILL.md ~/.claude/skills/dm-other-claudes/ ``` 2. Add a reference to this skill in your `CLAUDE.md` or project instructions so Claude knows how to use it. Example entry: ```markdown ### Cross-Claude Communication - Tool: `bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts` - Remote hosts: - `user@your-server-ip` — description of what runs there - Use `--skip-permissions` for autonomous tasks (no human on the remote end to approve) - Use `-s SESSION_ID` for follow-ups ``` 3. Verify SSH connectivity: ```bash ssh user@your-server "claude --version" ``` ## Usage ### Tool Location ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts [options] <user@host> "prompt" ``` ### New Conversation ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts user@server "check if the web server is running and healthy" ``` ### Continue Conversation (with session ID from previous call) ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts -s SESSION_ID user@server "now restart it" ``` ### Fully Autonomous Remote Task ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts --skip-permissions user@server "pull latest code and redeploy the service" ``` ### Options | Flag | Description | |------|-------------| | `-s, --session <id>` | Resume a previous session on the remote | | `-t, --timeout <sec>` | Timeout in seconds (default: 600 / 10min) | | `-f, --file <path>` | Read prompt from a file | | `--skip-permissions` | Add `--dangerously-skip-permissions` on remote (needed for autonomous tool use) | | `--max-turns <n>` | Limit how many tool-use turns the remote Claude can take | | `--model <model>` | Override the remote Claude's model | | `-v, --verbose` | Debug output | ## How to Use From the TUI (Instructions for Claude) When you (Claude in TUI) need to delegate work to a remote machine: ### Step 1: Send the initial task ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts --skip-permissions user@server "your task description here" ``` The tool outputs: - **stdout**: The remote Claude's response text - **stderr**: Metadata including `session_id` and cost ### Step 2: Process the response Read the response. If you need to follow up: ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts -s SESSION_ID user@server "follow-up question or next task" ``` ### Step 3: Report back to the user Once the collaboration is complete, summarize what happened. ### Multi-Turn Pattern For complex tasks that need back-and-forth: ```python # Pseudocode for the pattern response1 = dm-claude(host, "analyze the logs on this machine") # Parse response1, extract session_id response2 = dm-claude(host, session=sid, "now restart the failing service") # Parse response2 response3 = dm-claude(host, session=sid, "confirm it's healthy") # Report combined results to user ``` ## What the Remote Claude Sees The tool auto-prepends a preamble to the first message explaining: - This is a cross-Claude communication - The calling Claude is working with a human in a TUI - It should be thorough but concise - It should complete the task autonomously - Follow-ups may arrive via `--resume` On follow-up messages (`--session`), no preamble is added — the remote Claude already has the context from the resumed session. ## Important Notes 1. **Always use `--skip-permissions` for autonomous tasks** — without it, the remote Claude may not be able to use tools that require permission prompts (since there's no human on the remote end to approve) 2. **Use IP addresses if hostname resolution is unreliable** — if you have flaky DNS, use the IP directly (e.g., `[email protected]` instead of `user@myserver`) 3. **Timeout** — default is 10 minutes. For long tasks, increase with `-t 1200` or more 4. **Session IDs live on the remote machine** — they reference Claude sessions stored in `~/.claude/` on the remote host. The session ID only works with the same `user@host` 5. **Cost awareness** — each dm-claude call is a separate Claude API call on the remote machine. The tool reports cost in stderr 6. **The remote Claude has its own CLAUDE.md and skills** — it will follow its own instructions on that machine ## Common Patterns ### Delegate a task to a remote server ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts --skip-permissions user@server \ "Check the disk usage on /var and clean up any log files older than 7 days" ``` ### Check on a service ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts user@server \ "Is the API server running? Check with curl localhost:8080/health and show me the process" ``` ### Deploy code ```bash bun run ~/.claude/skills/dm-other-claudes/dm-claude.ts --skip-permissions user@server \ "cd ~/myapp && git pull && npm install && pm2 restart myapp" ``` ### Parallel work across multiple machines Use the Bash tool's `run_in_background` to send tasks to multiple machines simultaneously: ```bash # In parallel: bun run dm-claude.ts --skip-permissions user@server1 "task for server 1" bun run dm-claude.ts --skip-permissions user@server2 "task for server 2" ```

Preview in:

Security Status

Verified

Manually verified by security team

Time saved
How much time did this skill save you?

Related AI Tools

More Grow Business tools you might like

codex-collab

Free

Use when the user asks to invoke, delegate to, or collaborate with Codex on any task. Also use PROACTIVELY when an independent, non-Claude perspective from Codex would add value — second opinions on code, plans, architecture, or design decisions.

Rails Upgrade Analyzer

Free

Analyze Rails application upgrade path. Checks current version, finds latest release, fetches upgrade notes and diffs, then performs selective upgrade preserving local customizations.

Asta MCP — Academic Paper Search

Free

Domain expertise for Ai2 Asta MCP tools (Semantic Scholar corpus). Intent-to-tool routing, safe defaults, workflow patterns, and pitfall warnings for academic paper search, citation traversal, and author discovery.

Hand Drawn Diagrams

Free

Create hand-drawn Excalidraw diagrams, flows, explainers, wireframes, and page mockups. Default to monochrome sketch output; allow restrained color only for page mockups when the user explicitly wants webpage-like fidelity.

Move Code Quality Checker

Free

Analyzes Move language packages against the official Move Book Code Quality Checklist. Use this skill when reviewing Move code, checking Move 2024 Edition compliance, or analyzing Move packages for best practices. Activates automatically when working

Claude Memory Kit

Free

"Persistent memory system for Claude Code. Your agent remembers everything across sessions and projects. Two-layer architecture: hot cache (MEMORY.md) + knowledge wiki. Safety hooks prevent context loss. /close-day captures your day in one command. Z