meshguard 0.1.0

Python SDK for MeshGuard - Governance Control Plane for AI Agents

MeshGuard Python SDK

Python SDK for MeshGuard — Governance Control Plane for AI Agents.

Installation

pip install meshguard

With LangChain support:

pip install meshguard[langchain]

Quick Start

from meshguard import MeshGuardClient

# Initialize client
client = MeshGuardClient(
    gateway_url="https://dashboard.meshguard.app",
    agent_token="your-agent-token",
)

# Check if an action is allowed
decision = client.check("read:contacts")
if decision.allowed:
    print("Access granted!")
else:
    print(f"Denied: {decision.reason}")

# Enforce policy (raises PolicyDeniedError if denied)
client.enforce("read:contacts")

# Use context manager for governed code blocks
with client.govern("write:email") as decision:
    # This code only runs if allowed
    send_email(to="[e-mail]", body="Hello!")

Environment Variables

You can configure the client using environment variables:

export MESHGUARD_GATEWAY_URL="https://dashboard.meshguard.app"
export MESHGUARD_AGENT_TOKEN="your-agent-token"
export MESHGUARD_ADMIN_TOKEN="your-admin-token"  # For admin operations

Then simply:

from meshguard import MeshGuardClient

client = MeshGuardClient()  # Uses env vars

LangChain Integration

Govern Individual Tools

from meshguard import MeshGuardClient
from meshguard.langchain import governed_tool

client = MeshGuardClient()

@governed_tool("read:contacts", client=client)
def fetch_contacts(query: str) -> str:
    """Fetch contacts matching query."""
    return contacts_db.search(query)

# The tool only runs if policy allows "read:contacts"
result = fetch_contacts("John")

Wrap Existing Tools

from langchain.tools import DuckDuckGoSearchRun
from meshguard import MeshGuardClient
from meshguard.langchain import GovernedTool

client = MeshGuardClient()
search = DuckDuckGoSearchRun()

# Wrap the tool with governance
governed_search = GovernedTool(
    tool=search,
    action="read:web_search",
    client=client,
)

result = governed_search.run("MeshGuard AI governance")

Create Governed Agent

from langchain.llms import OpenAI
from langchain.agents import load_tools
from meshguard import MeshGuardClient
from meshguard.langchain import create_governed_agent

client = MeshGuardClient()
llm = OpenAI()
tools = load_tools(["serpapi", "llm-math"], llm=llm)

agent = create_governed_agent(
    llm=llm,
    tools=tools,
    client=client,
    action_map={
        "serpapi": "read:web_search",
        "Calculator": "execute:math",
    },
)

result = agent.run("What is 25 * 4?")

Handle Denied Actions

from meshguard import MeshGuardClient, PolicyDeniedError
from meshguard.langchain import governed_tool

client = MeshGuardClient()

def handle_denial(error, *args, **kwargs):
    return f"Sorry, I can't do that: {error.reason}"

@governed_tool("write:email", client=client, on_deny=handle_denial)
def send_email(to: str, body: str) -> str:
    # Send email...
    return "Email sent!"

# If denied, returns the denial message instead of raising
result = send_email("[e-mail]", "Hello!")

Admin Operations

from meshguard import MeshGuardClient

client = MeshGuardClient(admin_token="your-admin-token")

# List agents
agents = client.list_agents()
for agent in agents:
    print(f"{agent.name} ({agent.trust_tier})")

# Create agent
result = client.create_agent(
    name="my-agent",
    trust_tier="verified",
    tags=["production"],
)
print(f"Created agent: {result['id']}")
print(f"Token: {result['token']}")

# List policies
policies = client.list_policies()

# Get audit log
entries = client.get_audit_log(limit=10, decision="deny")

Proxy Requests

Route requests through MeshGuard for automatic governance:

from meshguard import MeshGuardClient

client = MeshGuardClient()

# GET request
response = client.get("/api/contacts", action="read:contacts")

# POST request
response = client.post(
    "/api/emails",
    action="write:email",
    json={"to": "[e-mail]", "body": "Hello!"},
)

Error Handling

from meshguard import (
    MeshGuardClient,
    MeshGuardError,
    AuthenticationError,
    PolicyDeniedError,
    RateLimitError,
)

client = MeshGuardClient()

try:
    client.enforce("delete:database")
except PolicyDeniedError as e:
    print(f"Action denied: {e.action}")
    print(f"Policy: {e.policy}")
    print(f"Reason: {e.reason}")
except AuthenticationError:
    print("Invalid or expired token")
except RateLimitError:
    print("Rate limit exceeded")
except MeshGuardError as e:
    print(f"MeshGuard error: {e}")

API Reference

MeshGuardClient

Method	Description
check(action)	Check if action is allowed (returns PolicyDecision)
enforce(action)	Enforce policy (raises PolicyDeniedError if denied)
govern(action)	Context manager for governed code blocks
health()	Check gateway health
list_agents()	List all agents (admin)
create_agent(name, trust_tier, tags)	Create agent (admin)
revoke_agent(agent_id)	Revoke agent (admin)
list_policies()	List policies (admin)
get_audit_log(limit, decision)	Get audit entries (admin)

LangChain Integration

Function/Class	Description
@governed_tool(action)	Decorator for governed tools
GovernedTool(tool, action)	Wrapper for existing tools
GovernedToolkit(tools, action_map)	Govern multiple tools
create_governed_agent(llm, tools)	Create governed agent

License

MIT License - see LICENSE for details.

Links

• Website: https://meshguard.app
• Dashboard: https://dashboard.meshguard.app
• Documentation: https://github.com/dbhurley/meshguard
• Issues: https://github.com/dbhurley/meshguard-python/issues