Claude IPCSAFE
AI-to-AI communication protocol for Claude, Gemini, and other AI assistants
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Enable AI-to-AI communication using simple natural language commands. Works with Claude Code, Gemini, ChatGPT, and any Python-capable AI assistant.
What It Does
Claude IPC MCP lets different AI assistants send messages to each other, even across different platforms and sessions. Think of it as email for AIs - simple, reliable, and persistent.
# AI #1 (Claude) Register this instance as claude Send message to gemini: Can you help with the database schema? # AI #2 (Gemini) Register this instance as gemini Check messages > "Can you help with the database schema?" - from claude
Quick Install (2 minutes)
# 1. Clone the repo git clone https://github.com/jdez427/claude-ipc-mcp.git cd claude-ipc-mcp # 2. Install UV package manager curl -LsSf https://astral.sh/uv/install.sh | sh # 3. Install dependencies uv sync # 4. For Claude Code: Run installer ./scripts/install-mcp.sh # 5. Restart Claude Code and test # Type: Register this instance as myname
Full installation guide: INSTALL.md
Key Features
- 💬 Natural language commands - No coding required
- 💾 Persistent messages - Messages survive restarts
- 🔄 Cross-platform - Works between different AI platforms
- 🎯 Simple setup - Install once, use everywhere
- 🔐 Optional security - Add authentication if needed
Basic Commands
Register this instance as alice # Set your name Send message to bob: Hello! # Send a message Check messages # Check inbox List instances # See who's online
Documentation
- INSTALL.md - Complete installation guide
- TROUBLESHOOTING.md - Common issues and solutions
- **[do
094e040b345dOBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add claude-ipc-mcp --env IPC_SHARED_SECRET=${IPC_SHARED_SECRET} -- uvx claude-ipc-mcp{
"mcpServers": {
"claude-ipc-mcp": {
"command": "uvx",
"args": [
"claude-ipc-mcp"
],
"env": {
"IPC_SHARED_SECRET": "${IPC_SHARED_SECRET}"
}
}
}
}Exposed tools (9)
6 read · 3 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
auto_process | read | Automatically check and process IPC messages (for use with auto-check feature) |
broadcast | read | Broadcast a message to all other Claude instances |
check | read | Check for new messages |
list_instances | read | List all active Claude instances |
register | read | Register this Claude instance with the IPC system |
rename | write | Rename your instance ID (rate limited to once per hour) |
send | write | Send a message to another Claude instance |
share_command | write | Execute a command and share output with another instance |
share_file | read | Share file content with another instance |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- declared (2 observation(s))
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (7)
.mcp.json.example
print(f"✅ IPC_SHARED_SECRET is set (length: {len(secret)})")curl -LsSf https://astral.sh/uv/install.sh | sh
curl -LsSf https://astral.sh/uv/install.sh | sh
curl -LsSf https://astral.sh/uv/install.sh | sh
curl -LsSf --insecure https://astral.sh/uv/install.sh | sh
curl -LsSf https://astral.sh/uv/install.sh | sh
Gates applied: no_behavioural_pass.
094e040b345dfull audit observations/trust-audit/mcp-server/jdez427__claude-ipc.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | 094e040b345d | SAFE | B | 89 | first audit |
Questions
What is the Claude IPC MCP server?
AI-to-AI communication protocol for Claude, Gemini, and other AI assistants
What tools does Claude IPC expose?
9 in total: 6 read-only, 3 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Claude IPC safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does Claude IPC need?
It reads IPC_SHARED_SECRET from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does Claude IPC run?
It speaks stdio, so it runs as a local process your client starts. It is published on PyPI as claude-ipc-mcp.
How current is this page?
The grade is for one exact copy of the source (094e040b345d), read on 2026-10-07. The repository is watched and re-audited when it changes.