v1.4.1

Full Changelog: v1.4.0...v1.4.1

v1.4.0 — Hermes Agent Bridge

What's New

Hermes Agent Bridge (NousResearch)

AI COMMS now connects to Hermes Agent — the self-improving AI agent by Nous Research with 40+ tools, skills, memory, cron scheduling, and multi-platform messaging.

Prefixes: !hermes / !ha

!hermes research competitor pricing and summarize in a table
!ha fix the deployment script and run tests

Hermes Agent exposes an OpenAI-compatible API server on port 8642. AI COMMS talks to it directly via /v1/chat/completions — no custom bridge server needed.

Setup

Hermes side (~/.hermes/.env):

API_SERVER_ENABLED=true
API_SERVER_KEY=your-secret-key

Then: hermes gateway

AI COMMS side (.env):

HERMES_API_URL=http://127.0.0.1:8642
HERMES_API_KEY=your-secret-key

6 Bridges

Bridge	Prefix	Port
Copilot (VS Code)	!copilot / !cp	3120
Claude Code	!claude / !cc	3121
Codex	!codex / !cx	3122
Cursor	!cursor / !cu	3123
OpenClaw	!claw / !oc	3124
Hermes Agent	!hermes / !ha	8642

Stats

• 126 tests, 0 failures
• 5 files changed, 177 insertions

v1.3.0 — Security Hardening

Security Release

7 security fixes addressing OWASP-class vulnerabilities found during codebase audit.

Fixes

#	Fix	File	Severity
1	SSRF guard — reject private/loopback IPs in agent URL validation	multi-agent.js	HIGH
2	JSON schema validation — validate AI-generated group commands before execution	orchestrator.js	CRITICAL
3	Remote agent sandbox — \execSync\ → \execFileSync\ (no shell), metachar detection, command whitelist	remote-agent.js	CRITICAL
4	Hub rate limiting — per-IP sliding-window on HTTP (60/min) + WS connect rate (10/min)	hub/server.js	HIGH
5	KDF swap — SHA-256 → PBKDF2 (100k iterations) for encryption key derivation	encryption.js	MEDIUM
6	Atomic audit log — crash-safe rotation via temp-file writes	audit-log.js	MEDIUM
7	Silent catches → audit log — blocked messages now logged with context for observability	orchestrator.js, admin.js	HIGH

Breaking Changes

• SSRF guard: Private/loopback agent URLs are now blocked by default. Set \ALLOW_PRIVATE_AGENTS=true\ if running agents on localhost.
• Encryption KDF: Keys derived from \SECURITY_ENCRYPTION_KEY\ now use PBKDF2 instead of SHA-256. Existing encrypted messages from v1.2.x will need re-encryption.
• Remote agent: Shell metacharacters (&&, |, \) in commands are now blocked. Only whitelisted executables (node, npm, git, python, etc.) are allowed.

Tests

8 new security tests (123 total, 0 failures)

Full Changelog: v1.2.8...v1.3.0

v1.2.8 — README: All 5 IDE Bridges

Updated README to reflect all 5 IDE bridge adapters (Copilot, Claude Code, Codex, Cursor, OpenClaw) instead of VS Code/Copilot-only language. Architecture diagram, intro, Agent Hub, and Multi-Agent sections now accurately describe the full bridge support.

Full Changelog: v1.2.7...v1.2.8

v1.2.7 — OpenClaw Hub Connector

What's New

OpenClaw Hub Connector — Bridge an OpenClaw Gateway to the AI COMMS Agent Hub mesh for cross-network agent discovery.

Changes

• src/openclaw-hub-connector.js — Dual WebSocket bridge: registers as openclaw agent on the Hub, forwards tasks/messages between OpenClaw Gateway and Hub agents
• Auto-reconnect (5s), heartbeat (25s), graceful shutdown
• New npm script: npm run connector:openclaw
• 5 new tests (115 total, 0 failures)
• Updated .env.example with connector env vars
• README: new OpenClaw Hub Connector section with architecture diagram

Architecture

OpenClaw Gateway ↔ Hub Connector ↔ Agent Hub ↔ Other Agents

Quick Start

export HUB_SECRET=your-secret
npm run connector:openclaw

Full Changelog: v1.2.6...v1.2.7

v1.2.6 — README cleanup + npm publish

v1.2.6

• Removed duplicated README section (389 lines of stale content)
• npm publish with clean README
• 5 bridges: Copilot, Claude Code, Codex, Cursor, OpenClaw
• CLI: doctor, --format, --verbose, structured exit codes
• 110 tests, 0 failures

Full Changelog: v1.2.5...v1.2.6

v1.2.5 — OpenClaw Bridge Adapter

What's New in v1.2.5

OpenClaw Bridge Adapter

AI COMMS now has 5 bridge adapters. The new OpenClaw bridge connects your WhatsApp/Telegram messages to a running OpenClaw Gateway (356k stars personal AI assistant).

Prefix	Port	Target
!copilot / !cp	3120	VS Code Copilot
!claude / !cc	3121	Claude Code CLI
!codex / !cx	3122	OpenAI Codex CLI
!cursor / !cu	3123	Cursor IDE
!claw / !oc	3124	OpenClaw Gateway (NEW)

Usage

ai-comms --bridge openclaw "summarize today's messages"

From WhatsApp/Telegram:

!claw ship checklist for the release
!oc summarize today's messages and draft a reply

Setup

1. Install OpenClaw: npm install -g openclaw@latest
2. Start the gateway: openclaw gateway --port 18789
3. Start the bridge adapter on port 3124
4. Set OPENCLAW_BRIDGE_PORT=3124 in .env

Tests

110 total, 0 failures (+3 new OpenClaw bridge tests).

Full Changelog: v1.2.4...v1.2.5

v1.2.4 — Doctor, Exit Codes, Format & Verbose

What's New in v1.2.4

doctor command

Run ai-comms doctor to diagnose your setup -- checks Node.js version, .env file, AI provider config, API connectivity, all 4 bridges, hub server, and npm.

Structured exit codes (sysexits.h)

Code	Name	Meaning
0	OK	Success
2	USAGE	Bad usage
66	NOINPUT	Input missing
69	UNAVAILABLE	Service offline
77	NOPERM	Auth failure
78	CONFIG	Config error

--format flag

-f json, -f csv, -f table for agents status, doctor, and more. Pipe-friendly structured output.

--verbose flag

-v enables verbose stderr logging. -V is now version (was -v).

Tests

107 total, 0 failures (+12 new tests for exit codes, format, verbose, doctor parsing).

Full Changelog: v1.2.3...v1.2.4

v1.2.3 — CLI with Full Computer Control

Highlights

CLI with full computer control — AI COMMS now has a terminal interface. Run ai-comms for an interactive REPL with 12 native tools (file ops, shell commands, code search, HTTP requests), or relay tasks to any running IDE bridge. No WhatsApp or messaging platform needed.

---

What's New

CLI — Standalone Mode

• 12 native tools: read_file, write_file, append_file, list_directory, search_files, grep, run_command, file_info, move_file, delete_file, http_request, system_info
• AI tool-calling loop — the AI decides which tools to use, executes them, and iterates
• Works with any provider that supports function calling (OpenAI, Groq, DeepSeek, Mistral, xAI, Copilot, etc.)
• Falls back to simple chat for providers without tool support

CLI — Bridge Relay Mode

• Route tasks to any running bridge: ai-comms --bridge copilot "fix the tests"
• Supports all 4 bridges: Copilot, Claude Code, Codex, Cursor
• Token auth, 3-minute timeout, same contract as WhatsApp/Telegram

CLI — Interactive REPL

• Color-coded prompt with conversation history
• Commands: /bridge, /bridges, /tools, /provider, /clear, /help, /quit
• Agent management: ai-comms agents status shows all bridge statuses

CLI — One-Shot Mode

• ai-comms "your task here" — runs, prints result, exits
• Perfect for scripts, CI/CD pipelines, and automation

Global Install

npm install -g ai-comms
ai-comms --help

Tests

• 12 new CLI tool tests (95 total, 0 failures)
• Covers: file read/write/delete, directory listing, command execution, system info, grep, search, HTTP

---

Full Changelog: v1.2.2...v1.2.3

Full Changelog: v1.2.2...v1.2.3

v1.2.2 — Claude Code, Codex & Cursor Bridges

Highlights

4 IDE Bridges — AI COMMS now connects to Copilot (VS Code), Claude Code, OpenAI Codex, and Cursor from WhatsApp and Telegram. Each bridge uses the same HTTP contract — send a task from your phone, get results from any IDE agent.

---

What's New

Claude Code Bridge

• New bridge client: src/claude-code-bridge.js
• WhatsApp/Telegram prefix: !claude or !cc
• Default port: 3121
• Token auth: CLAUDE_CODE_BRIDGE_TOKEN
• 3-minute timeout for multi-step coding tasks

Codex Bridge

• New bridge client: src/codex-bridge.js
• WhatsApp/Telegram prefix: !codex or !cx
• Default port: 3122
• Token auth: CODEX_BRIDGE_TOKEN

Cursor Bridge

• New bridge client: src/cursor-bridge.js
• WhatsApp/Telegram prefix: !cursor or !cu
• Default port: 3123
• Token auth: CURSOR_BRIDGE_TOKEN

Bridge Architecture

All 4 bridges share the same HTTP contract — any server that implements these endpoints can plug in:

GET  /health          → 200 OK
POST /chat            → { "message": "...", "sender": "..." }
                      ← { "response": "..." }

Tests

• 9 new bridge tests (83 total, 0 failures)
• Prefix detection, availability checks, error handling for all 3 new bridges

---

New Environment Variables

Variable	Default	Description
CLAUDE_CODE_BRIDGE_PORT	3121	Claude Code bridge server port
CLAUDE_CODE_BRIDGE_TOKEN	—	Auth token for Claude Code bridge
CODEX_BRIDGE_PORT	3122	Codex bridge server port
CODEX_BRIDGE_TOKEN	—	Auth token for Codex bridge
CURSOR_BRIDGE_PORT	3123	Cursor bridge server port
CURSOR_BRIDGE_TOKEN	—	Auth token for Cursor bridge

---

Full Changelog: v1.2.1...v1.2.2