Runlayer is an enterprise AI platform that helps companies move from early AI experiments to broad, governed adoption of AI agents. It sits between an organisation’s employees, AI clients, agents and tools, and gives security teams one place to see, approve and control how agents connect to business systems.
It is for companies whose staff already use AI clients and agents faster than security can review them, and who want that use to grow under control instead of being blocked.
At A Glance
| Vendor | Runlayer (Anysource Inc.), United States |
|---|---|
| Category | AI control plane, MCP gateway, agent identity and governance |
| Deployment | Hosted, or self-hosted in the customer’s AWS account |
| Identities covered | Users, groups, roles, agent accounts and delegated users |
| Agentic AI | MCP gateway, agent IAM, runtime security, governed agent builder |
| Customers | Lemonade, Gusto, PagerDuty, AppLovin, Opendoor |
| Certifications | SOC 2, HIPAA, GDPR |
| Funding | Series A, Felicis and Khosla Ventures |
| Market classification | Agentic AI Identity; also AI Safety & Guardrails |
What Runlayer Does
An MCP gateway on its own only covers part of the problem, so Runlayer wraps one with a catalogue, identity controls and observability. It pairs control for the security team with enablement for employees: the same platform that decides what an agent may reach also gives staff a vetted set of tools and a way to build agents with the right permissions. Usage and audit history then show whether an AI programme is paying off, while every agent action stays accountable to a person and a policy.
Products
Runlayer MCP Gateway
The gateway is the path between AI clients and approved MCP servers. Every call passes through policy, OAuth and runtime security checks and lands in an audit trail, which gives security teams a single enforcement point for agent traffic.
Runtime Security
Runtime security protects against what agents read and do. It detects prompt injection and invisible characters, credential and PII exposure, data exfiltration, intent drift and agent manipulation, as well as unnecessary tool access and risky skill files. Teams can start in alert-only mode to set a baseline and move to block mode once they trust the rules.
Runlayer Agents
Runlayer Agents is a builder for governed agents, with native Slack integration. Each agent runs with a scoped identity, so business teams can automate work without handing an agent broad standing access.
Build Tools And Artifacts
For developers, Runlayer Deploy hosts custom MCP servers, alongside a Hooks SDK and a Platform API. Artifacts let users create, find and share reports, interactive pages and files produced with AI.
Runlayer Catalog
Runlayer Catalog is an internal registry for distributing approved connectors, skills and plugins, so employees pick from a vetted set instead of installing tools on their own.
Shadow AI And Observability
Discovery finds unmanaged MCP servers, skills, plugins and AI clients on managed devices through the existing MDM, and observability tracks usage, spend and adoption across the organisation.
How It Works
Runlayer runs as a hosted deployment or self-hosted in the customer’s own AWS account, either on ECS on Fargate with Terraform or through the Runlayer Operator on a customer-managed EKS or GKE cluster.
The platform connects to the identity provider, approved tools, MCP servers and existing AI clients. Agent traffic flows from the client through the gateway to the MCP server, and the policy engine evaluates each request against users, groups, roles, agent accounts, tools, OAuth state, client, network and session context. Agents can also be managed as code with Terraform, with zero trust networking for Runlayer agents.
Identities And Systems Covered
| Agent accounts | Each request is tied to an actor, agent account, delegated user, credential source, policy decision and audit record |
|---|---|
| Autonomous agents | Machine-to-machine tokens |
| Delegated agents | On-behalf-of tokens that preserve user context when an agent acts for a specific person |
| Human users | Users, groups and roles from the identity provider |
| Credentials | Session grants, personal or shared, decide which OAuth credentials an agent account can use |
Agentic AI Capabilities
Agent Identity And Delegation
For delegated calls, Runlayer enforces the intersection of the agent account’s policies, the user’s policies and server-level policies, so an agent can never do more than both the person and the server allow. The audit trail records user, agent account, client, connector, tool, policy result, credential source and outcome separately for each action.
Human Approval
Policies can pause risky actions until a person approves them, combining the runtime scanners with a human check before the call proceeds.
Integrations
| Area | Integrations |
|---|---|
| AI clients | Claude, Claude Code, Cursor, ChatGPT, Codex, VS Code, Windsurf, Goose |
| MCP servers | Catalogue servers in 15 categories, including GitHub, Atlassian, Slack, Notion, Linear, Google Calendar, Sentry, Google Drive, Snowflake, Salesforce and Workday |
| Security and identity | Okta MCP server for users and groups; CrowdStrike Falcon for hosts, detections and incidents |
| Infrastructure | Terraform; existing MDM for discovery |
Company Background
| Legal entity | Anysource Inc. |
|---|---|
| Founders | Andy Berman, Tal Peretz and Vitor Balocco |
| Funding | A reported $30M Series A from Felicis and Khosla Ventures in June 2026, bringing total funding to $42M |
| Locations | New York, with remote staff in the US |
Note: This overview is compiled from Runlayer’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.