TrueFoundry review

TrueFoundry combines an AI gateway with a platform for deploying and managing models, agents, and AI workloads. Its product materials describe deployment on-premises, in a private cloud environment (VPC), in hybrid setups, across public clouds, and in air-gapped environments.

The gateway side covers three areas: LLM traffic, agent connections, and MCP connections. The LLM gateway offers unified model routing, fallbacks, rate limits, cost controls, and observability. The MCP gateway is for discovering and securing connections between agents, tools, and MCP servers. The Agent Gateway covers agent identity, access controls, usage policies, and tracing.

TrueFoundry also describes observability for prompts, tool and model execution, metrics, latency, and outcomes, alongside audit logging, role-based access controls, compliance features, and runtime guardrails. In short, it targets organizations that want deployment flexibility and centralized controls, not just a proxy for model API calls. The breadth of that pitch is useful, but how much of it a team needs depends on its deployment and governance requirements.

Deployment choice is the clearest strength. TrueFoundry says teams can run it on-premises, in a VPC, in air-gapped environments, or across multiple clouds. That gives organizations options when a public-cloud-only gateway does not fit their infrastructure or data-residency requirements. Enterprise lists VPC, on-premises, and data residency as included capabilities, though the pricing page does not publish a price for that plan.

The product also brings more than model routing into its scope. Its feature list includes fallbacks, rate limits, cost controls, and observability for LLM requests, plus identity and policy controls for agents and connections to MCP servers. Audit logs, role-based access controls, and runtime guardrails add governance features around those workflows. The pricing page lists SSO and custom roles on Pro, too.

That combination may suit teams trying to manage models and agent-related access from one platform rather than assembling separate controls. But the feature list is not evidence of how completely it governs every agent action. A Reddit commenter cautions that local tools and subprocesses may not pass through a gateway, so teams should check which actions their architecture actually routes through TrueFoundry.

Per-user pricing can make Pro harder to budget as a team grows. According to the pricing page, Pro costs $25 per user per month and includes 20,000 requests and tool calls per user each month, up to 10 teams, and 30-day log retention. The page lists a $20 overage for each additional 100,000 requests and tool calls. Model-provider charges are separate pass-through costs, and no annual Pro price is published.

The free Developer plan is capped at three users, 10,000 requests and tool calls per user per month, two teams, and seven-day log retention. Enterprise pricing requires a custom annual commitment; its request allowance and retention are custom. Agent Gateway and Agent Harness also have custom pricing, and third-party or custom guardrails are Enterprise-only and billed by provider usage.

Public user evidence is thin. One Reddit thread says its author could not find much production experience with TrueFoundry beyond the first few months; another commenter described the developer experience as clunky. Those are individual reports, not a broad verdict, but they leave less public evidence to assess than the feature list might suggest.

The pricing page lists Developer at $0, Pro at $25 per user/month, and Enterprise at a custom annual commitment. Developer includes up to 3 users, 10K requests and tool calls per user/month, 2 teams, and 7-day log retention. Pro removes the user cap and raises the allowance to 20K requests and tool calls per user/month, with up to 10 teams and 30-day retention. That is a meaningful jump in capacity, but the per-user price can make costs harder to predict as a team grows. The page lists Pro overage at $20 per additional 100K requests and tool calls.

Model-provider charges are pass-through and billed separately. Third-party and custom guardrails are Enterprise-only and billed by provider usage. Agent Gateway and Agent Harness have custom pricing. Enterprise also requires a custom annual commitment; no annual Pro price is published. Budget beyond the seat price, especially if usage, guardrails, or agent products are central to your setup.

Treat a move as a gateway and deployment project, not just a model-provider change. Map your model routes, fallbacks, rate limits, agent and tool connections, access policies, and observability needs before switching. TrueFoundry lists VPC, on-premises, air-gapped, hybrid, and multi-cloud deployment options, but the published material does not document migration tools, supported import formats, or whether logs and traces can be exported. Confirm those details before planning a cutover or exit.

To reduce risk, inventory existing prompts, routing rules, credentials, guardrails, and audit requirements, then test representative workloads in parallel. Keep a rollback path until routing and policy behavior are verified. We cannot say what tends to make customers leave based on Reddit: no Reddit posts or comments were supplied. The material also does not establish how portable TrueFoundry-specific agent, gateway, or observability configurations are, so ask for export and offboarding terms in writing.