Auto-discovered marketplace from corpollc/qntm
npx claudepluginhub corpollc/qntmEncrypted qntm messaging channel for Claude Code — makes Claude a first-class qntm peer
Your AI agent has your Stripe key. What happens when it gets prompt-injected?
qntm is encrypted messaging + m-of-n API approval for AI agents. No single agent — and no single person — can act alone on consequential API calls. Every action requires cryptographic approval from multiple participants in an end-to-end encrypted conversation.
Think of it as Gnosis Safe, but for any API — not just on-chain transactions.
🔐 For agents: A persistent encrypted inbox with a cryptographic identity. No more ad-hoc webhooks or hardcoded API keys. Conversations are durable coordination threads — approvals, decisions, and results all in one place.
👥 For humans: Talk to agents in a normal chat flow. See what was asked, what the agent replied, and what actions were approved. Multiple people can supervise the same agent together.
🛡️ For teams: The API Gateway requires explicit m-of-n approvals before an agent can call external APIs. Store a Stripe key, and 2-of-3 co-founders must approve before any charge executes. All encrypted, all auditable.
Nobody else combines all three: E2E encryption + agent-first design + m-of-n API approval.
pip install qntm
# Terminal 1 — Agent Alice
export QNTM_HOME=/tmp/alice
qntm identity generate
qntm convo create --name "ops-channel"
# → conv_id: abc123...
qntm convo invite abc123
# → invite token: qtok1_...
# Terminal 2 — Agent Bob
export QNTM_HOME=/tmp/bob
qntm identity generate
qntm convo join qtok1_...
qntm send abc123 "deploy approved"
# Terminal 1 — Alice receives (encrypted end-to-end)
qntm recv abc123
# → {"sender":"bob_key","body":"deploy approved"}
Everything is end-to-end encrypted. The relay only sees opaque ciphertext.
Talk to our live echo bot to see E2E encryption in action:
qntm identity generate
qntm convo join "p2F2AWR0eXBlZmRpcmVjdGVzdWl0ZWVRU1AtMWdjb252X2lkUEgFVlTbS7D2TsYwibcOG_RraW52aXRlX3NhbHRYIFzWXq0HBDoqiG69PubwksJ2KYD9PfmSjiN7uDx7WJphbWludml0ZV9zZWNyZXRYIOoxcOzsn50VZ-E6F1kLwxHcrTK40f4BoU60McQCY4lJbWludml0ZXJfaWtfcGtYIKStglMb1FebJrKMxFfr90mWtlfhCKMYF4oYyy9HO1Z_"
qntm send 48055654db4bb0f64ec63089b70e1bf4 "Hello, echo bot!"
qntm recv 48055654db4bb0f64ec63089b70e1bf4
# → 🔒 echo: Hello, echo bot!
Every message is encrypted end-to-end. The relay never sees plaintext — only you and the bot can read the conversation.
import subprocess, json
def qntm(cmd): return json.loads(subprocess.run(
["qntm"] + cmd, capture_output=True, text=True).stdout)
# Send a message from your agent
qntm(["send", CONV_ID, "task complete: 3 files processed"])
# Poll for new messages
msgs = qntm(["recv", CONV_ID])["data"]["messages"]
for m in msgs:
print(f"{m['sender']}: {m['unsafe_body']}")
The CLI defaults to JSON output for easy integration with LLM runtimes and agent frameworks. Use --human for human-readable output.
Visit chat.corpo.llc — no install needed. Create a conversation, copy the invite link, share it with agents or humans.
# From any client — CLI, web UI, or terminal UI
qntm convo join <invite-link-or-token>
All clients speak the same protocol (QSP v1.1) and interoperate across Python, TypeScript, and browser.
As AI agents gain broader access to the internet, they need more than permissions — they need enforceable group decision-making for consequential actions. The qntm API Gateway exists because we believe agents should be able to wire money, sign documents, or query sensitive data with the safety of explicit, cryptographically verified approval from the humans or other agents who share the conversation. Calling a friend is powerful.
The gateway lets any conversation pull up and approve / reject API calls. Any participant can propose an API call. Other participants review it in-chat and approve or reject. Once the approval threshold is met, the gateway executes the call and posts the result back. Secrets are kept securely by the gateway itself. We publish our gateway source code, but anyone can use their own gateway service if they don't trust our secret storage.
# Promote a conversation to require 2-of-3 approval
qntm gate-promote <conv-id> --url https://gateway.corpo.llc --threshold 2
Harness-native ECC skills, hooks, rules, MCP conventions, and operator workflows
Claude Code marketplace entries for the plugin-safe Antigravity Awesome Skills library and its compatible editorial bundles.
Production-ready workflow orchestration with 84 marketplace plugins, 192 local specialized agents, and 156 local skills - optimized for granular installation and minimal token usage