output — skytale
Every agent protocol sends plaintext through servers that can read it. Skytale encrypts all of them. ── The Problem ────────────────────────────── without skytale: Agent A → Server (reads everything) → Agent B with skytale: Agent A → Relay (sees nothing) → Agent B ── Protocols ──────────────────────────────── PROTOCOL TODAY WITH SKYTALE ──────────────── ────────────────── ──────────────── A2A (Google) TLS transport only MLS end-to-end MCP (Anthropic) stdio / HTTP MLS end-to-end ACP (IBM) HTTP REST MLS end-to-end SLIM (AGNTCY) pub/sub MLS end-to-end ANP / LMOS / NLIP HTTPS MLS end-to-end ── Features ───────────────────────────────── Zero-Knowledge Relay — routes ciphertext, sees nothing Verified Memory — SharedContext, cryptographic attribution Cryptographic ID — DID:key per agent, every message signed Audit Trail — Hash-chained logs, W3C Trace Context AGNTCY/SLIM compatible · Linux Foundation · RFC 9420 Cross-protocol bridge: A2A ↔ ACP ↔ MCP ↔ SLIM ───────────────────────────────────────────── $ pip install skytale-sdk [SIGNUP] GitHub OAuth, 10 seconds free · 1M messages/mo · no credit card
── Frameworks ───────────────────────────── langgraph $ pip install skytale-sdk[langgraph] crewai $ pip install skytale-sdk[crewai] mcp $ pip install skytale-sdk[mcp] openai-agents $ pip install skytale-sdk[openai] pydantic-ai $ pip install skytale-sdk[pydantic] smolagents $ pip install skytale-sdk[smolagents] agno $ pip install skytale-sdk[agno] google-adk $ pip install skytale-sdk[adk] autogen $ pip install skytale-sdk[autogen] llamaindex $ pip install skytale-sdk[llamaindex] strands $ pip install skytale-sdk[strands] agentkit $ pip install skytale-sdk[agentkit] elizaos $ npm install @skytalesh/plugin-eliza goat-sdk $ npm install @skytalesh/plugin-goat mastra $ npm install @skytalesh/sdk ── Protocol Adapters ────────────────────── SLIM, A2A, ACP, MCP, ANP, LMOS, NLIP
── quickstart.py ────────────────────────── $ cat quickstart.py from skytale_sdk import SkytaleChannelManager # Create an encrypted channel mgr = SkytaleChannelManager( identity=b"agent-a", api_key="sk_live_..." ) mgr.create("acme/trading/signals") # Send — MLS encrypted before leaving your process mgr.send("acme/trading/signals", "whale alert: ETH accumulation detected") # Receive — decrypted only by channel members msgs = mgr.receive("acme/trading/signals") # Relay saw: a4f8c2e1 9b3d7a06 5c8f1e4d... # Agent sees: "whale alert: ETH accumulation" $ python quickstart.py
── Risks ────────────────────────────────── [CRITICAL] Compromised agent poisons your state Without attribution, can't tell which agent wrote false data. Downstream agents trust everything. ref: OWASP Agentic Top 10 ASI06 fix: Every write signed with author's DID. [CRITICAL] Can't explain agent decision chain Regulators ask for the chain. No trace linking. ref: EU AI Act Article 12 fix: Hash-chained logs + W3C Trace Context. [WARNING] Two agents overwrite each other Without causal ordering, one write silently wins. fix: HLC ordering + conflict resolution. [WARNING] Agent contradicts yesterday Without persistent state, decisions remade each session. fix: SharedContext with cryptographic attribution. ── Compliance ───────────────────────────── [PASS] EU AI Act — Article 12, record-keeping [PASS] NIST AI 600-1 — agent identity & delegation [PASS] OWASP Top 10 — agent hijacking & data leakage 3/3 compliance checks passed
── Plans ────────────────────────────────── free $0/mo 1M messages/mo Unlimited agents, keys, channels MLS encryption, federation, compliance [SIGNUP] no credit card pro $99/mo (coming soon) Unlimited messages, 99.9% SLA ── Enterprise ───────────────────────────── enterprise custom (coming soon) Dedicated relays, on-prem, 99.99% contact@skytale.sh ───────────────────────────────────────────── Self-host: Apache 2.0 github.com/nicholasraimbault/skytale
── Cryptography Sandbox ─────────────────── Live cryptography sandbox Real MLS (RFC 9420) running in your browser via WebAssembly. $ encrypt "your message here" MLS encrypt via mls-rs compiled to WASM. Real ciphertext. $ decrypt Decrypts the last encrypted message. $ keygen Generates an Ed25519 keypair + DID:key identifier. $ hash "any data" SHA-256 hash. Same algorithm Skytale uses for audit log chaining. ───────────────────────────────────────────── Try it: type a command below. Example: $ encrypt "whale alert: ETH accumulation detected"
Loading...
Loading...
Loading...
Loading...
Get your agents encrypted. ── Install ────────────────────────────────── $ pip install skytale-sdk $ export SKYTALE_API_KEY=<your-api-key> [SIGNUP] [COPY SETUP] ── Add to your code ───────────────────────
from skytale_sdk.integrations import langgraph
app = langgraph.encrypt(
graph.compile(),
channel="myorg/myapp/agents",
)
[COPY CODE]Run your agents. Check the activity tab. 15 frameworks supported → ── AI coding tools ──────────────────────── $ claude mcp add skytale -- \ [COPY] python -m skytale_sdk.integrations._mcp Then ask your AI: "Create an encrypted channel"