For the builders of agentic AI
Give enterprises the confidence to let agents do more.
Maya adds independent, per-agent control on the network path.
No traffic decryption. No agent code changes.
research-agent
A defined mandate
MAYA Per-agent policy
Other agents keep running.
01 · The missing layer
One agent in a fleet goes wrong.
Can you identify it? Can you stop it?
Agents connect to models, data, tools, and other agents. A single workflow can cross runtimes and cloud boundaries. Enterprises need to know which agent made each connection and whether that connection belongs within its mandate.
Maya makes the logical agent a governable unit on the network path. Security teams can allow, limit, or stop an agent, or individual flows within its traffic, while the rest of the fleet keeps working.
Identity
Defines the principal’s permitted access.
Runtime controls
Govern activity within the agent’s runtime.
Maya
Enforces the agent’s mandate on its connections.
An independent layer that complements model and runtime safeguards.
02 · From intent to control
You set the policy. Maya applies it on the wire.
Define an agent’s policy mandate once, centrally.
Discover agents and attribute each connection to the agent that made it.
Expose undeclared agents and traffic that deviates from an agent’s mandate.
Evaluate the observed behavior against the policy you defined.
Allow, limit, or stop the agent or the specific flow. Enforcement is reversible.
03 · An independent control layer
The agent can act. The enterprise sets the boundaries.
04 · One governed fleet
Available now on AWS and GCP, in your cloud account.
Where agents run
Remote workers, Linux servers, Kubernetes, cloud agent runtimes, and container platforms.
Where they connect
LLM endpoints, databases, object storage, MCP servers, and SaaS services.
How teams operate
Central policy and connection evidence, with OpenTelemetry export and gRPC / REST APIs. Works with identity providers including Okta, SPIFFE, and cloud IAM.
05 · The people behind Maya
Founder & CEO
datapath · identity
Networking datapath architect with nearly three decades building forwarding planes, deep packet inspection, QoS, and control-plane security at Cisco, Juniper, and Brocade, and cloud infrastructure at IBM and Oracle. He built Maya's eBPF datapath and agent-identity model.
Nine US patents. M.S. Computer Engineering, Santa Clara University.
Co-founder & CTO
security · product
Founding engineer of Cisco's network infrastructure security group, where he worked on Control Plane Policing and deep packet inspection, then product and go-to-market leader at SentinelOne, Cyberinc (acquired by Forcepoint), Elastic Security, and Anomali.
Published IETF RFC author. MBA, UC Berkeley Haas.
VP of Engineering
cloud · production · compliance
Most recently Director of Software Engineering at Palo Alto Networks, running the data platform behind Strata Cloud Manager and multiple cloud-delivered security services: 2 trillion events and 6 PB a day in more than 20 countries, at 99.99% uptime. Twenty-five years in networking, cloud infrastructure, and security, going back to deep packet inspection and control-plane protection in Cisco IOS. She has taken products through federal certification including FedRAMP and FIPS 140-2, working directly with the validation labs. She owns Maya's engineering organization and Maya's path to production: control plane, telemetry, multi-cloud deployment, and release.
First-named inventor on a US patent for management-plane architecture. B.Tech, College of Engineering Trivandrum.
06 · Build with us
Now onboarding design and research partners.
We’re looking for model builders, agent platforms, and research teams to evaluate how independent network enforcement can support enterprise agent deployments.
Bring a workflow that matters to your customers. Together, we can test how agent identity, policy, and connection evidence hold up as autonomy grows.
A proposed first collaboration
Agree on the runtime, destinations, and behavior to evaluate.
Define a mandate, then test undeclared agents and connections beyond it.
Assess agent attribution, enforcement, and the effect on the remaining fleet.
Let’s start a conversation
What would your agents do
with a boundary you could trust?
Or email [email protected] to explore a partnership.
Research inquiries: [email protected]
Cookies
We use analytics cookies to see how this site is used so we can make it better. Nothing is stored until you say yes. See our privacy policy.
