Built an open AI Gateway for routing/observing LLM + MCP + Agent traffic — feedback welcome

Hey all — wanted to share something we’ve been building in case it’s useful to anyone working with LLMs/agents, and also get feedback from this community.

Disclosure: I work at TrueFoundry, so take this as a “here’s what we made,” not a neutral recommendation.

The problem we kept running into: once you’re calling multiple LLM providers, spinning up MCP servers/tools, and running agents in production, you lose visibility fast — no unified logging, no cost tracking per token/user, no easy way to add guardrails across providers, no failover when a provider goes down.

So we built an AI Gateway that sits in front of all of it:

  • One OpenAI-compatible API for any provider (OpenAI, Anthropic, etc.) with automatic fallback/load balancing
  • An MCP Gateway layer for governing which tools/agents can call what
  • Token-level cost attribution and budget limits
  • Cross-provider guardrails (not locked to one vendor’s safety layer)
  • Full observability — traces, logs, prompt analytics
  • Can run as SaaS, on-prem, or fully air-gapped (SOC 2 / HIPAA compliant, which matters if you’re in healthcare/finance)

Docs/repo: TrueFoundry · GitHub

Curious if others here building agent-based projects have run into the same pain points — how are you currently handling multi-provider routing or cost tracking in your own projects?

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.