Developers
Build agents you can trust in production.
Write your agent with any framework. MAQPNA checks every tool and model call against your policy, asks a person when a rule says so, and records the result. The same code runs on your laptop and in the cluster.
1. Install the CLI
curl -fsSL https://maqpna.com/install.sh | shirm https://maqpna.com/install.ps1 | iex2. Run it locally
Three commands. A governed agent.
maqpna dev up starts the identity broker, the gateway with a starter policy, an audit ledger and a test tool server on your machine.
maqpna dev up
local MAQPNA is up (/home/you/project/.maqpna)
gateway http://127.0.0.1:8080
broker http://127.0.0.1:8081
mcp echo -> http://127.0.0.1:8090/mcp
next: maqpna dev run -- <your agent command>
maqpna dev run -- python agent.py
maqpna dev timeline --last
TIME KIND SERVER/TOOL DECISION DETAIL
14:05:09 tool_call echo/echo allow matched rule read-only-tools
14:05:11 tool_call echo/delete_resource deny system namespaces are off-limits to agents3. Add the SDK
Python and TypeScript. Or plain HTTP.
The SDKs read the session token MAQPNA gives your agent and turn decisions into typed results. Any MCP or HTTP client works too.
| Language | Install | Docs |
|---|---|---|
| Python 3.10+ | pip install kosa8-maqpna | Python SDK |
| TypeScript, Node 20.3+ | npm install @kosa8/maqpna | TypeScript SDK |
| Framework adapters | pip install "kosa8-maqpna[langgraph]"also langchain, openai-agents, claude-agent-sdk, crewai | Adapters |
| Go | Not published yet. Ask us for access. | Go SDK |
| Any language | HTTP and MCP to the gateway | Gateway API |
- MCP
- A2A
- LangGraph
- LangChain
- OpenAI Agents SDK
- Claude Agent SDK
- CrewAI
- Vercel AI SDK
- OpenAI-compatible models
Developer guide
Step by step, from first call to deploy.
Tool calls over MCP
Call tools through the gateway and handle allow, deny and hold.
Model calls
Send LLM calls through the gateway with policy, DLP and budgets.
Approvals and decisions
Wait for a person without blocking your agent loop.
Agent to agent
Delegate to other agents over A2A with scoped identities.
Test agents and policies
Policy tests in CI, replay, and fork to compare runs.
Package and deploy
From
maqpna devto an Agent resource on Kubernetes.
Tutorials
Build something real, from an empty folder.
Support agent with DLP and approvals
Redact customer data and hold refunds for a person.
A governed LangGraph agent
Put an existing graph behind the gateway.
Multi-agent over A2A
A planner and workers, each with its own identity.
A cost-controlled LLM app
Hard budgets per session, user and tenant.
Policy as code in CI
Lint, test and promote policies with Git.
A private coding agent
A local model and a microVM sandbox, nothing leaves the machine.
Reference
Every endpoint, field and command.
Gateway API
Tool, model and session endpoints.
Admin API
Approvals, audit, policy and kill switch.
Kubernetes resources
Agent, ToolPolicy, BudgetPolicy and more.
CLI commands
Every
maqpnacommand and flag.
Kosa8 Labs stack
Add local models and microVMs. Keep one audit trail.
MAQPNA works with any OpenAI-compatible model and any Kubernetes sandbox. With llamay and kosa8, also from Kosa8 Labs, each session gets its own model context and a microVM you can fork.
Do I need Kubernetes to develop?
No. maqpna dev up runs on macOS, Linux and Windows. You need Kubernetes only to run agents in production.
Do I have to use the SDK?
No. Any MCP client or HTTP client can call the gateway. The SDKs add typed results and handle approvals for you.
Do I need an API key or a GPU?
No. The local setup includes a test tool server, and llamay can serve a small model on a laptop CPU.
Does my code change between laptop and cluster?
No. The SDKs read the same environment in both places. You move policies and agents as Kubernetes resources.
Your first governed call in five minutes.
No Kubernetes, GPU or API key. One command starts a local MAQPNA.
curl -fsSL https://maqpna.com/install.sh | shirm https://maqpna.com/install.ps1 | iex