Opencontroller

What it does

Opencontroller takes in the AI agents an organization runs across clouds, Kubernetes clusters, vendor platforms and developer laptops, including agents that were never registered. It builds a live inventory of those agents, each with a named owner and identity, and applies policy decisions, budget checks, guardrails, traces and release gates to the requests they make to models, tools and APIs.

Teams typically begin with read-only discovery, then add owners, release gates and runtime rules as agents move toward production. Requests pass a single checkpoint that checks identity, policy and remaining budget before a call reaches a model, tool or API, and records the trace against an owner and cost. Agents stay on the runtimes they already use, since the product sits in the request path.

It suits enterprises that already run agents across several clouds and need governance for security, cost and compliance teams. Access is through a demo request and sales contact. Deployment can run in the customer's VPC, on-premises or air-gapped. Governance starts at $100K a year, and model tokens are billed by the model provider, not Lyzr.

What it can do

• Read-only discovery of agents across AWS, Azure, Google Cloud, Kubernetes and developer laptops
• Per-agent registry with owner, purpose, tools, versions and risk tier
• Agent identity issued through the customer's identity provider, with no shared API keys
• Release gates from multi-turn, red-team and regression evals before promotion
• Agent Gateway enforcing caller authentication, rate limits and access rules on each call
• LLM Gateway with virtual keys, team budgets, model routing and fallback
• Guardrails for prompt injection, personal data, secrets and per-tool limits
• OpenTelemetry traces with cost and owner on each span

Works with

• Amazon Bedrock AgentCore
• Gemini Enterprise Agent Platform
• Microsoft Foundry
• Agentforce
• Okta (identity provider in example deployment)

Before you choose

• Pricing starts at $100K a year for governance; model token charges are paid to the model provider directly.
• Deployment and identity integrations depend on the customer's existing cloud, identity provider and CI/CD setup.
• Setup is described as taking less than a week for a first environment, based on the vendor's own claim.

Listed under

• Coding
• MCP and agent tools

Get started

No self-serve sign-up or install path is listed on the page.

1. Connect one environment for read-only discovery across clouds, clusters and laptops.
2. Assign owners and identities to discovered agents and register them in the registry.
3. Enable gateways, guardrails and release gates on the agents that reach production.

Priced in US dollars, billed yearly. Model tokens are billed by your model provider, not Lyzr. The free Build tier covers agent building, not Opencontroller.

Source: lyzr.ai/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• OpenController: Unified AI Agent Control Plane | Lyzr AI ↗www.lyzr.ai
• Lyzr Pricing | Enterprise AI Agent Platform ↗www.lyzr.ai
• Lyzr | Take your AI agents to production, faster. ↗www.lyzr.ai

Company website

lyzr.ai: product, pricing and policy pages

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.