Opencontroller by lyzr
Lyzr AI · Productivity · Business · Other
Opencontroller by Lyzr is an agent control plane that finds every AI agent across clouds, laptops and vendor platforms, then enforces identity, policy and budget on each call it makes.
About Opencontroller by lyzr
What Is Opencontroller by Lyzr
Opencontroller is the governance product from Lyzr, an enterprise AI infrastructure company based in Jersey City. It targets a problem most IT teams now recognize: agents get built in every department, each one holds its own API keys, and nobody can say how many are running or what they cost. The control plane puts one inventory, one owner list and one policy model over all of them.
It works as a layer in the request path. Before an agent reaches a model, an MCP tool, an enterprise API or another agent, Opencontroller checks who is calling, whether the call is allowed and whether budget remains, then records the result. Traffic keeps flowing to the destination you already use. Lyzr states that control-plane data, credentials and telemetry do not need to reach its own servers in a customer-managed deployment.
The scope is deliberately narrow. Opencontroller does not build agents, and it does not replace your identity provider, CI/CD pipeline or SIEM. It expects those systems to stay in place and connects to them through OIDC, SAML, SCIM, MCP, A2A and OpenTelemetry. Teams that only run one agent in one cloud will find the setup heavier than the problem. The product pays off when ten or fifty agents spread across Bedrock, Foundry, Vertex and a few developer laptops need a common set of rules.
Getting Started
- Request a demo and pick one environment to connect, since discovery is read-only and changes nothing in your cloud.
- Run a discovery scan across cloud accounts, Kubernetes clusters and developer laptops to build the AI agent inventory.
- Assign a named owner, an identity and a risk tier to each agent that shows up in the registry.
- Add evals and a release gate, so versions that fail task success, groundedness or red-team checks stay out of production.
- Route agent calls through the Agent Gateway and LLM Gateway, then set spend limits, guardrails and traces for each owner.
Product Information
A quick look at Opencontroller by lyzr's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Platform and security engineers
- CIOs and heads of AI
- Compliance and risk teams
Tasks
- Building an AI agent inventory
- Enforcing AI agent governance on every call
- Tracking AI agent spend management
- Retiring shared API keys
Scenarios
- A company running agents on three clouds after acquisitions or departmental pilots
- A regulated team that cannot send telemetry outside its network
- Coding agents on developer laptops
Key features
Read-only discovery across clouds and laptops
The first module scans AWS accounts, Azure subscriptions, Google Cloud projects, Kubernetes clusters and developer machines, then lists what it finds as registered or shadow agents. Scans are read-only, so a team can start without touching running workloads. Lyzr says a first environment is typically connected in under a week.
Agent registry with owners and identities
Every agent gets one record: purpose, runtime, tools, versions, risk tier and a named human owner. Alongside it, the agent receives its own identity issued through Okta or another OIDC provider, with scopes such as read access to cases or refunds below a set amount. Shared keys drop to zero, and each credential carries an expiry.
Release gates backed by evals
Multi-turn, red-team and regression evals run before a version reaches production. A failing score on task success or groundedness blocks promotion to staging, and the failing cases go back to the owner. Releases support canary rollout, two-person approval and one-step rollback, while your existing CI pipeline keeps building.
Agent Gateway and LLM Gateway
The Agent Gateway decides who can call each agent, whether that caller is a public app, an internal service or another agent over A2A, and applies rate limits per caller. The LLM Gateway adds virtual keys, team budgets, routing and fallback across model providers, so a rate-limited request can fail over to a second model without a code change.
Guardrails and spend limits
Inputs and outputs are checked in both directions. Prompt injection attempts are blocked, personal data and card numbers are redacted, and tool calls above a threshold, such as a refund over $2,500, wait for human approval. Per-tool limits apply to MCP servers as well as to hosted models.
Traces plus a refinement loop
Agent observability runs on OpenTelemetry, with spans for the agent run, gateway authorization, guardrails, model calls and tool calls, each carrying a cost figure. Failures are grouped into clusters, a proposed fix is tested against held-out cases, and the corrected behavior becomes a new eval case.
Pros and cons
Pros
- Runs inside your own network, including air-gapped setups, so credentials and telemetry stay under your control.
- Covers agents built on other platforms, with adapters for Bedrock AgentCore, Microsoft Foundry and Gemini Enterprise, plus MCP and A2A support.
- Puts policy in the request path, which means a blocked call never reaches the model or the tool.
- Ten modules can be adopted one at a time, so a team can start with discovery and add gates later.
- Setup is measured in days rather than quarters, and the read-only first scan lowers the political cost of getting started.
Cons
- Pricing starts at $100,000 a year, which rules it out for small teams and single-agent projects.
- There is no self-serve signup and no public price list beyond the starting figure, so every deal goes through sales.
- The value depends on having agents in production across more than one runtime, and a team with one agent on one cloud will not use most of the modules.
- Day-to-day operation still needs a platform owner who maintains policies, eval datasets and owner records.
Frequently asked questions
It discovers AI agents across your clouds and laptops, gives each one an owner and an identity, and enforces policy, budget and guardrails before any call reaches a model, tool or API. The registry, releases and traces all sit in the same console.
Related content
Explore related tools, skills, and articles for Opencontroller by lyzr.
Opencontroller by lyzr Alternatives
Base Code
Base44 (Wix) · Coding · ProductivityBase Code is a shared cloud development environment from Base44. Teammates describe changes to an existing GitHub codebase in chat, preview them live and send them back as pull requests.
Knocket
Knocket (Tencent RTC team) · Productivity · Business · ChatbotKnocket is a free live chat and contact page tool that lets visitors message you from your website, app, or a shareable link, with every reply handled from Telegram or one inbox.
FloorGen
Vinit Kumar (FloorGen) · Productivity · China AIFloorGen is a browser-based floor plan maker that turns a sentence or a sketch into an editable 2D plan and a 3D walkthrough. Export to PDF, DXF or video, and draw without an account.