TrueFoundry is an enterprise AI infrastructure company that helps platform and security teams run LLMs, MCP servers and AI agents under one set of controls. Its gateways sit between applications, the models they call and the tools agents use, handling routing, access, guardrails and tracing in one place.
The problem it solves is sprawl. Once a company has several model providers, a set of MCP servers and a growing number of agents, keys, costs and permissions end up scattered across teams. TrueFoundry brings them back under central control while keeping data inside the customer’s own infrastructure.
At A Glance
| Vendor | TrueFoundry (Ensemble Labs Inc) |
|---|---|
| Category | AI gateway and AI infrastructure platform |
| Identities covered | AI agents, agent service accounts, human users, MCP tool tokens |
| Deployment | On-premises, VPC, hybrid, air-gapped, multi-cloud or public cloud |
| Agentic AI | Agent Gateway, MCP Gateway, Agent Skills Registry, TrueForge |
| Certifications | SOC 2 |
| Customers | NVIDIA, NetApp, Innovaccer, Sportsbet, Aviva Credito |
| Market classification | AI Engineering & Infrastructure |
What TrueFoundry Does
TrueFoundry’s approach is a single governed layer in front of every model, tool and agent. Every request passes through a gateway that decides where it goes, who is allowed to make it and what it may touch, and records it in an immutable audit log. Because agents, MCP servers and models share that one layer, a single policy and audit trail covers the whole chain from prompt to tool call, and the same platform can also host and tune the models themselves.
Products
AI Gateway
The AI Gateway is for teams that call several model providers and want one API, one bill of controls and no single point of failure. It connects to OpenAI, Claude, Gemini, Groq, Mistral and other LLMs, routes by cost, latency, availability and task, and falls back to a secondary model automatically when a request fails. TrueFailover extends this to whole-provider outages, rerouting enterprise AI traffic automatically. Rate limits can be set for each user, service or endpoint, and guardrails filter PII and detect toxic content. The gateway adds a reported sub-3ms of internal latency.
MCP Gateway
The MCP Gateway gives agents and developers one place to discover and use MCP servers, offering “unified access to MCP Servers via a secure, discoverable, and enterprise-grade MCP gateway.” It ships prebuilt servers for common enterprise tools, lets teams register any internal or proprietary API as an MCP server, and links both self-hosted and vendor-run servers. Each server gets its own access policy and audit trail.
Agent Gateway
The Agent Gateway addresses the agent sprawl problem: knowing which agents exist, what they can do and what they cost. It keeps a central registry of framework-agnostic agents, sets quotas, budgets and access boundaries across environments, and captures end-to-end traces across agent steps, model calls and tool calls. It also restricts who can deploy, execute or modify an agent and applies policy to every tool invocation. It is in pre-launch.
TrueForge
TrueForge is an open-source, vendor-neutral agent harness for teams that want to build and run managed agents without tying themselves to one model provider.
Model Hosting, Prompts And Skills
The AI Deployment Platform and Training and Fine-Tuning host, serve and tune models on the customer’s own infrastructure, so the platform that governs model traffic can also run the models. Prompt Management and the Agent Skills Registry give teams a shared home for the prompts and agent skills they reuse across projects.
How It Works
Applications and agents send their requests to a single Gateway Base URL, the endpoint for the AI Gateway service. Self-hosted installations get their own base URL inside the customer’s environment. The platform can be deployed on-premises, in a VPC, hybrid, air-gapped, across several clouds or on public cloud.
Every tool call that crosses the gateway is checked against OAuth2, RBAC and metadata policies before it reaches the target, and traces leave the platform in OpenTelemetry-compatible form. Platform monitoring covers latency, token usage, cost and GPU utilisation.
Identities And Systems Covered
| AI agents | Authentication, identity and service account management for agents is centralised at the gateway layer |
|---|---|
| Agent frameworks | LangChain, CrewAI, AutoGen and custom agents |
| Human users | Federated login with SSO and role-based access control |
| Tool access tokens | OAuth 2.0 with dynamic discovery secures the tokens agents use to reach MCP servers |
Integrations
| Area | Integrations |
|---|---|
| Model providers | OpenAI, Claude, Gemini, Groq, Mistral, AWS Bedrock |
| Guardrails | OpenAI Moderation, AWS Guardrails, Azure Content Safety, Azure PII detection |
| Identity providers | Okta, Azure AD and other IdPs |
| MCP tools | Slack, Confluence, Sentry, Datadog |
| Observability | Grafana, Prometheus, OpenTelemetry |
Company Background
| Legal entity and headquarters | Ensemble Labs Inc, San Francisco, United States. |
|---|---|
| Leadership | Nikunj Bajaj, Co-Founder and CEO. |
| Acquisitions | Seldon AI (June 2026). |
| Recognition | AIBoomi Awards 2026, DevTools and Infrastructure category. |
Note: This overview is compiled from TrueFoundry’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.