lockstock 0.1.0

The TCP/IP of AI Agency - Cryptographic identity and memory for AI agents

LockStock Python SDK

The TCP/IP of AI Agency

Cryptographic identity and memory continuity for autonomous AI agents.

What is LockStock?

LockStock provides three critical capabilities for AI agents:

1. The Lock (Security): Cryptographic identity verification - agents can't be impersonated
2. The Stock (Memory): Portable state - agents never forget, even across server migrations
3. The Two Smoking Barrels (Proof): Unforgeable audit trail - prove what happened, when, in order

Installation

pip install lockstock

Quick Start

1. Bootstrap a New Agent

from lockstock import LockStockAgent, Task

# Create agent
agent = LockStockAgent(
    server_url="https://lockstock.example.com",
    client_id="trading-bot-42",
    secret="your-agent-secret-key"
)

# Bootstrap (first time only)
agent.bootstrap()

# Agent is now ready with cryptographic identity

2. Perform Actions

# Authenticate each action
response = agent.authenticate(Task.QUERY)

if response.accepted:
    print(f"✅ Authenticated at sequence #{agent.passport.last_sequence}")
    print(f"State hash: {response.state_hash}")
    print(f"Server timestamp: {response.server_timestamp}")

3. Agent Teleportation (Zero-Downtime Migration)

import json

# On Server A: Export passport
passport = agent.export_passport()
with open("passport.json", "w") as f:
    json.dump(passport, f)

# Upload to S3, Redis, database, anywhere...

# On Server B: Import passport
with open("passport.json", "r") as f:
    passport = json.load(f)

agent = LockStockAgent.from_passport(
    passport=passport,
    server_url="https://new-server.example.com",
    secret="your-agent-secret-key"
)

# Agent continues from EXACTLY where it left off
# Zero memory loss. Zero downtime.

4. View Action History (The Golden Thread)

# Get complete audit trail
timeline = agent.get_action_timeline()
for action in timeline:
    print(action)

# Output:
# Seq #1: BOOTSTRAP (hash: a3f4b2e1...)
# Seq #2: QUERY (hash: b2e1c9d7...)
# Seq #3: EXECUTE (hash: c9d7a4f8...)

Use Cases

Multi-Cloud Migration

# AWS
passport = agent.export_passport()
upload_to_s3("my-bucket", "agent-passport.json", passport)

# Agent crashes on AWS

# Azure
passport = download_from_s3("my-bucket", "agent-passport.json")
agent = LockStockAgent.from_passport(passport, server_url="azure-endpoint", secret=secret)

# Agent resumes at exactly the same sequence number
# No memory loss. Cryptographic proof of continuity.

Hot Standby (High Availability)

# Primary agent (active)
while True:
    passport = primary_agent.export_passport()
    redis.set("agent-42-passport", json.dumps(passport))
    primary_agent.authenticate(Task.QUERY)
    time.sleep(1)

# Standby agent (passive, different server)
while True:
    if not is_primary_alive():
        # Take over instantly
        passport_data = redis.get("agent-42-passport")
        passport = json.loads(passport_data)
        standby_agent = LockStockAgent.from_passport(passport, server_url, secret)
        standby_agent.authenticate(Task.EXECUTE)
        break
    time.sleep(1)

LangChain Integration

from langchain.agents import AgentExecutor
from lockstock import LockStockAgent, Task

# Wrap your LangChain agent with LockStock
lockstock_agent = LockStockAgent(
    server_url="https://lockstock.example.com",
    client_id="research-agent",
    secret="agent-secret"
)
lockstock_agent.bootstrap()

# Your LangChain agent now has:
# - Cryptographic identity
# - Persistent memory across restarts
# - Clone detection
# - Audit trail

executor = AgentExecutor(agent=your_agent, tools=tools)

# Before each action, authenticate
lockstock_agent.authenticate(Task.QUERY)
result = executor.invoke({"input": "Research AAPL stock"})

# Export passport for crash recovery
passport = lockstock_agent.export_passport()
save_passport(passport)

Error Handling

from lockstock.exceptions import (
    CircuitBreakerError,
    SplitBrainError,
    MemoryGapError,
    AuthenticationError,
)

try:
    response = agent.authenticate(Task.EXECUTE)
except CircuitBreakerError as e:
    # Agent is making requests too fast (rogue behavior detected)
    print(f"🚨 LOCKED: {e}")
    # Alert security team, investigate
except SplitBrainError as e:
    # Two instances detected (clone attack)
    print(f"🧠 CLONE DETECTED: {e}")
    # Emergency shutdown
except MemoryGapError as e:
    # Agent has a gap in sequence (data corruption)
    print(f"⚠️ MEMORY GAP: {e}")
    # Restore from backup
except AuthenticationError as e:
    # Invalid signature or unauthorized task
    print(f"❌ AUTH FAILED: {e}")

Features

Identity = The Sum of History

Traditional auth: "I have the API key" ✅ Authenticated

LockStock: "I have the API key + cryptographic proof of my entire action history" ✅ Authenticated + Verified

Clone Detection (Split Brain Prevention)

If an attacker clones your agent:

Server: ✅ Primary agent authenticated at Seq #50
Server: ❌ Clone rejected - parent hash #49 already spent
        "Split brain detected. Original agent alerted."

Velocity Monitoring

Built-in circuit breaker detects rogue agents:

# Normal behavior: 1 req/sec
agent.authenticate(Task.QUERY)  # ✅ Accepted

# Rogue behavior: 100 req/sec (infinite loop)
for i in range(100):
    agent.authenticate(Task.EXECUTE)  # ❌ Circuit breaker trips

# CircuitBreakerError raised
# Agent locked. Security team alerted.

API Reference

LockStockAgent

Constructor:

agent = LockStockAgent(
    server_url: str,
    client_id: str,
    secret: str,
    passport: Optional[AgentPassport] = None
)

Methods:

• bootstrap() - Initialize new agent identity
• authenticate(task: Task) -> VerifyResponse - Authenticate action
• export_passport() -> dict - Export for migration
• get_action_timeline() -> List[str] - View history
• get_status() -> dict - Get agent status

Class Methods:

• from_passport(passport, server_url, secret) -> LockStockAgent - Resume from passport

Task (Enum)

Available task types:

• Task.BOOTSTRAP
• Task.DEPLOY
• Task.RESTART
• Task.BACKUP
• Task.UPDATE
• Task.QUERY
• Task.EXECUTE

AgentPassport

The agent's cryptographic "soul":

@dataclass
class AgentPassport:
    agent_id: str
    last_sequence: int
    last_hash: str
    state_matrix: Matrix
    last_server_timestamp: int
    action_history: List[ActionRecord]

Architecture

LockStock uses:

• Braid Group B₃ over finite field F₆₅₅₃₇ for topological state verification
• HMAC-SHA256 for signature verification
• Holder-of-Key authentication (must have full state matrix)
• Dual-Timestamp Architecture (logical sequence + physical server timestamp)
• DAG-based Linear Chain with parent hash validation

Comparison

Feature	Session IDs	JWT	LockStock
Continuity	❌ None	❌ Just identity	✅ Full history
Clone Detection	❌ No	❌ No	✅ Yes
Portability	❌ Lost on restart	❌ Not designed for it	✅ Export/import
Time Dependency	✅ No	❌ Yes (exp claim)	✅ No
Memory	❌ External DB	❌ Stateless	✅ Cryptographic

Support

• Documentation: https://docs.lockstock.dev
• Issues: https://github.com/lockstock/lockstock/issues
• Community: https://discord.gg/lockstock

License

MIT License - see LICENSE file for details.

---

Built with LockStock. The whole nine yards. 🔒📦💨