TL;DR
- Helicone is open source and ships two related things: a proxy-based observability platform (change one URL, not your SDK, and get full request/response logging, cost, latency, and quality tracking) and a separate, newer open-source AI Gateway repo covering routing across 20+ providers.
- Mintlify acquired Helicone on March 3, 2026; the product is in maintenance mode now, meaning bug fixes and new model support continue but active feature work has stopped, and Mintlify says it will help existing customers migrate.
What Helicone actually is
The integration model is the part I think is genuinely well designed: instead of wrapping your LLM client in a new SDK, you change one URL. Every request passes through Helicone's proxy, which logs the full request and response, tracks token counts and cost, and applies whatever behavior you've configured - caching, rate limiting, custom metadata - before forwarding the call to the actual provider. For debugging agent or RAG pipelines specifically, it gives you trace and session inspection across the whole call chain, not just a single request/response pair.
Separately, Helicone maintains its own open-source AI Gateway - one API in front of OpenAI, Anthropic, Google, Bedrock, and 20+ other providers, with smart routing that's aware of provider uptime and your own rate limits, aiming to always hit the fastest, cheapest, or most reliable option depending on how you've configured it. Both the observability platform and the gateway are self-hostable via Docker or Helm if you don't want to run on Helicone's cloud.
The acquisition, and what "maintenance mode" actually covers
Mintlify announced the acquisition on March 3, 2026. Per Helicone's own post about it, the product moves into maintenance mode: security updates, bug fixes, and new model support continue to ship, but active feature development has ended, and Mintlify says it will work with existing customers on migrating elsewhere. That's a meaningfully different status from either "actively developed" or "shut down" - it's closer to "stable, but the roadmap stopped moving on the date of acquisition."
Practically, that means: if you're already running Helicone, nothing breaks tomorrow, and you'll keep getting security patches and new model additions. If you're choosing new observability or gateway infrastructure today, you're choosing something whose feature roadmap has already ended, which is worth weighing against actively developed alternatives before committing not because what exists today is bad, but because the gap between it and a competitor with continued investment only grows from here.
Where it's still genuinely good, frozen roadmap and all
The proxy-based integration, the caching and rate limiting bundled into the same request path, and the cost/latency/quality tracking are all still solid mechanisms today - none of that stops working because feature development paused. If what you need is exactly what's already built - a drop-in observability layer without an SDK rewrite - maintenance mode doesn't change whether it does that job well right now.
Where it matters more is if your needs are going to grow: more sophisticated governance, deeper agent tracing, new provider integrations beyond what's already planned. One dev.to writeup covers the broader 2026 LLM proxy landscape shift this acquisition is part of, and there's a hands-on tool review worth reading if you want a working engineer's take on it before the maintenance-mode news, to see what people liked about it independent of this development.
Where I'd point you if you're picking new infrastructure today
The one-line proxy pattern Helicone popularized - point your base URL at the gateway instead of rewriting your client is the same integration model TrueFoundry's AI Gateway uses, so switching that specific habit isn't the hard part.
Where the two diverge is what's behind that URL: TrueFoundry's metrics dashboard covers the same cost, latency, and quality ground Helicone's observability product does, and adds RBAC, per-team budgets, and PII/prompt-injection guardrails as part of an actively developed product rather than one whose feature roadmap stopped on acquisition day. That's the actual tradeoff if you're choosing today: proven, frozen functionality versus a comparable starting point that's still moving.
Key Features of TrueFoundry
AI Gateway: Acts as a centralized control plane, enabling secure and efficient communication between applications and AI models. It supports features like unified API interfaces, access control, rate limiting, and observability.
MCP Servers: Facilitate the interaction between AI agents and external tools or services. TrueFoundry allows the deployment of dedicated MCP servers to manage agent traffic, enforce policies, and scale model access.
Tracing: Provides end-to-end observability by tracing every step from user input to agent response. This includes tracking user prompts, system messages, tool inputs, LLM calls, and workflow decisions.
Security & Compliance: Ensures data protection and compliance with standards like SOC 2, HIPAA, and GDPR. Features include role-based access control (RBAC), audit logging, and runtime security measures.
Deployment Flexibility: Supports various deployment options, including on-premises, virtual private cloud (VPC), air-gapped, hybrid, or public cloud environments, providing complete control over data and infrastructure.
TrueFoundry is a leading enterprise platform because it unifies AI deployment, observability, and governance in one scalable solution. Its advanced features, like the AI Gateway, MCP servers, and end-to-end tracing, give organizations full control, security, and transparency, making it ideal for managing complex AI applications at scale.
If you're currently on Helicone, has the maintenance-mode news actually changed your plans, or does it do everything you need already and the frozen roadmap just doesn't matter for your use case?
Top comments (0)