TrueFoundry
Also known as: TrueFoundry AI Gateway
Enterprise AI gateway and MCP gateway unifying 250 plus models with routing, guardrails, observability, and governance across SaaS, VPC, and air gapped deployments.
TrueFoundry provides an enterprise grade AI gateway that unifies access to more than 250 models from providers such as OpenAI, Anthropic, Google, Groq, and Mistral through a single OpenAI compatible API, with centralized key management, intelligent routing, failover, guardrails, rate limiting, observability, and flexible deployment. It sits directly in the production inference path as a control plane between applications and model providers, managing routing, governance, security, and cost without requiring changes to application logic. The company is independent, recognized in Gartner platform engineering research, and has expanded its control plane through the acquisition of Seldon AI.
The gateway enforces role based access control, single sign on, OAuth, and policy controls including PII and toxicity guardrails across all LLM traffic, and provides ClickHouse backed analytics for cost and usage breakdowns by model, team, user, or metadata. Reliability features include latency based and weighted routing, health based failover, semantic caching that can cut redundant calls, and support for both hosted and self hosted open source models such as Llama and Mistral through vLLM, SGLang, KServe, and Triton. Reported scale reaches billions of requests a month with multi region and high uptime targets.
A distinguishing capability for agentic use cases is a native MCP gateway. It exposes virtual MCP servers that aggregate tools like database queries, API calls, and file access behind one managed endpoint, injects OAuth identity so agents act as specific users with user level permissions on every tool call, and supports hybrid deployment where sensitive tools stay on premise while reasoning happens in the cloud. Deployment spans fully managed SaaS, customer cloud, VPC, on premise, and air gapped, with data residency controls. Pricing includes a free Developer tier for prototyping, a Pro tier from $499 a month for small teams shipping real features with usage based limits, and a sales led Enterprise tier for strict data controls and advanced governance, with a self hosted gateway plane running roughly $600 a month in cloud infrastructure.
Vendor details
Canonical URL
https://www.truefoundry.com
Category
Agent infrastructure
Subcategory
LLM gateway and routing
Funding status
Independent. Recognized in Gartner platform engineering research. Expanded its enterprise AI control plane through the acquisition of Seldon AI. Reports enterprise scale of billions of requests a month.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
One OpenAI compatible API to 250 plus models across OpenAI, Anthropic, Google, Groq, and Mistral, with routing, failover, caching, and guardrails. Native MCP gateway with virtual MCP servers and OAuth identity injection connects tools like Slack, GitHub, Confluence, and Datadog. Compatible with LangChain and OpenAI SDKs, and serves self hosted models via vLLM, SGLang, KServe, and Triton.
In practice
Your enterprise needs one governed endpoint for every provider with RBAC and audit. TrueFoundry gives a single OpenAI compatible API across 250 plus models, enforcing single sign on, quotas, and PII guardrails on all traffic.
Your agents call sensitive internal tools and you cannot share superuser credentials. The MCP gateway injects OAuth identity so each agent acts as a specific user with permissions enforced on every tool call.
Regulatory rules require sensitive data to stay on premise. You run the gateway plane in your VPC or air gapped, keep data local, and still route reasoning to cloud models under one policy layer.
Sources & related URLs
Agentic Index coverage score
9.0 / 14 capabilities · 64%
| Integrations & Tool Calling250 plus models, tool and function calling, native MCP tool access, docs 2026-07-06 | Full |
|---|---|
| Workflow OrchestrationMulti step agent workflow support via the MCP gateway, docs 2026-07-06 | Partial |
| Knowledge Grounding & RAGEmbeddings, reranking, and RAG components in the gateway, docs 2026-07-06 | Partial |
| Human Oversight & GuardrailsPII and toxicity guardrails and policy enforcement as a first class control, docs 2026-07-06 | Full |
| Security, Identity & GovernanceRBAC, SSO, OAuth, policy controls, air gapped support, docs 2026-07-06 | Full |
| Observability & AuditabilityClickHouse backed cost and usage analytics and monitoring, docs 2026-07-06 | Full |
| Memory & State PersistenceMCP gateway manages agent memory, docs 2026-07-06 | Partial |
| Deployment & Data ResidencySaaS, VPC, on premise, and air gapped with residency controls, docs 2026-07-06 | Full |
| Prebuilt Agents, Templates & PacksNo prebuilt agent templates library | Unable to verify |
| Triggers & Channel CoverageNo event triggers or channel coverage | Unable to verify |
| Model Flexibility & RoutingRouting and failover across 250 plus models plus self hosted inference, docs 2026-07-06 | Full |
| APIs, SDKs & MCP ExtensibilityOpenAI compatible API, SDKs, and native MCP gateway, docs 2026-07-06 | Full |
| Testing, Debugging & OptimizationPlayground, model comparison, and prompt management | Partial |
| Browser & Computer UseNo browser or computer use | Unable to verify |
The Agentic Index coverage score grades every vendor Full, Partial or Unable to verify against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
TrueFoundry introduced native human-in-the-loop approval workflows for Model Context Protocol (MCP) tools directly at the AI Gateway boundary. The gateway intercepts designated tool calls, notifies an approver, and suspends the execution state until authorization is manually granted or denied.
Bears on: MCP / tool calling / API
View sourceTrueFoundry announced the availability of its AI Gateway on VMware Cloud Foundation (VCF) and vSphere Kubernetes Service. This allows organizations to deploy TrueFoundry's unified model access, policy enforcement, and logging entirely within their own private cloud infrastructure.
Bears on: Security / enterprise
View sourceTrueFoundry introduced Ask TFY, a conversational interface built directly into its AI Gateway that acts as an AI agent for platform operations. The interface allows engineering teams to use natural language to query real-time configuration data, analyze traces, troubleshoot bugs, and generate YAML configuration updates.
Bears on: Observability / auditability
View sourcePricing
Free Developer tier; Pro from $499/mo; Enterprise contact sales
usage (requests and users) plus tier
Included quota
Developer tier is free for prototyping and testing. Pro from $499 a month unlocks higher limits, better performance, and essential governance, with additional usage billed at per unit rates. Tiers are set by usage, number of signed up users, and depth of offering such as AI Gateway and MCP Gateway.
What is public
The pricing page shows three tiers (Developer free, Pro from $499 a month, Enterprise) with usage based limits, but exact usage rates and Enterprise figures are gated to FAQs and sales.
Billing mechanics
Fully managed SaaS has no hosting cost beyond optional data storage on your own infrastructure. Self hosting the gateway plane or both planes adds roughly $600 to $1,000 a month in cloud infrastructure. Tiers scale by usage, signed up users, and the depth of the offering (AI Gateway, MCP Gateway).
Cost watchouts
Usage add ons above Pro limits and self hosted infrastructure costs are not fully visible on the pricing page. Enterprise governance, air gapped deployment, and premium SLAs require a sales conversation.
Variable cost rationale
The Pro base is a visible monthly fee, but usage add ons on requests and users, plus self hosted gateway plane infrastructure of roughly $600 a month, mean workload volume drives most of the real cost.
Overage / add-ons
Pro and Enterprise bill additional usage at per unit rates above included limits; specifics are noted in FAQs and scoped with sales.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free Developer tier for prototyping; live sandbox with no card
Lowest paid plan
Pro from $499 a month with usage based limits plus per unit usage charges
Commercial notes
Independent. Recognized in Gartner platform engineering research. Expanded its control plane through the acquisition of Seldon AI. Reports enterprise scale of billions of requests a month.
Key ambiguities
Per unit usage rates above included limits and Enterprise pricing are not published and must be scoped with sales.
Related vendors
- Acrab — Singapore compute infrastructure company building a full stack…
- AgentOps — Agent observability and reliability platform with broad model and…
- Agno — High-performance agent runtime and framework (formerly Phidata) with…
- AIsa — Unified resource and payment gateway for AI agents that lets them…
- AlphaBitCore — AI control plane that governs how models, agents, tools, and…
- Anchor Browser — Cloud hosted browser infrastructure that lets AI agents operate real…
Alternatives to TrueFoundry
The closest documented capability profiles to TrueFoundry among agent infrastructure platforms tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Clawvisor8.5 / 14Adds documented Prebuilt Agents, Templates & Packs
- Braintrust8.0 / 14Fuller documented coverage on Testing, Debugging & Optimization
- Dome Systems8.0 / 14Adds documented Triggers & Channel Coverage
- LiteLLM7.0 / 14A lighter documented profile than TrueFoundryTrueFoundry vs LiteLLM →
- Netlify10.0 / 14Adds documented Prebuilt Agents, Templates & Packs and Triggers & Channel Coverage
- Portkey8.0 / 14Fuller documented coverage on Testing, Debugging & Optimization
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded