Helicone Got Acquired — Here's How to Choose Your Next LLM Cost Tool
On March 3, 2026, Mintlify acquired Helicone. The Helicone team moved to Mintlify's San Francisco office. The product is now in maintenance mode — security updates and bug fixes only, no new features.
If you're one of the 16,000+ organizations that used Helicone, you need to migrate. But to where?
This isn't a sales pitch disguised as a comparison. We built AISpendGuard and we'll be upfront about that. But we'll also tell you when a competitor is the better choice — because picking the wrong tool costs more than any subscription.
What You Lose When Helicone Goes to Maintenance Mode
Helicone was a proxy gateway — you changed your base URL and all API calls routed through Helicone's servers. This gave you:
- Request/response logging — full prompt and completion storage
- Cost tracking — per-call cost attribution
- Caching — Redis-based with configurable TTL (up to 95% cost reduction)
- Low latency — Rust-based gateway with 8ms P50
- 200+ LLM support
Maintenance mode means: no new provider support, no new features, no guarantees it stays online long-term. The Mintlify team is working with customers on migration — but that clock is ticking.
The Bigger Question: Do You Still Need a Proxy?
Before picking a replacement, ask yourself: what did you actually use Helicone for?
Most Helicone users fall into two camps:
Camp 1: "I need full observability"
You debug prompts, trace agent chains, manage prompt versions, run evaluations. You need to see the actual content of API calls.
Your best options: Langfuse or Braintrust.
Camp 2: "I need cost tracking"
You want to know where your AI money goes — which features cost what, which models are wasteful, whether your spend is trending up or down. You don't need to see prompts.
Your best option: a tags-only tool like AISpendGuard.
The distinction matters because proxy-based tools (which store your prompts) create obligations that tags-only tools don't — especially under GDPR.
Honest Comparison: 5 Helicone Alternatives
1. Langfuse (Best for Full Observability)
What it is: Open-source LLM tracing and observability platform. Acquired by ClickHouse in January 2026.
Pricing: Self-hosted free. Cloud: Free 50K observations/mo, Core $29/mo, Pro $199/mo, Enterprise $2,499/mo.
Why it's great:
- 20,000+ GitHub stars — the community standard
- Deep tracing for agents, RAG pipelines, and complex chains
- Prompt management with versioning
- Evaluations (LLM-as-a-judge, human labeling)
- Self-hosting option for full data control
Why you might hesitate:
- Self-hosting requires PostgreSQL + ClickHouse + Redis + Kubernetes
- Cloud version stores prompts (same privacy concern as Helicone)
- Also acquired — ClickHouse's priorities may shift Langfuse's roadmap
- Core tier ($29/mo) jumped to Pro ($199/mo) for SOC 2 and longer retention
Best for: Engineering teams that need tracing, evals, and prompt management. The closest 1:1 Helicone replacement for full observability.
2. Braintrust (Best for Enterprise Teams)
What it is: CI/CD evaluations and production monitoring. $80M Series B at $800M valuation (February 2026).
Pricing: Starter $0/mo (usage-based: $4/GB storage, $2.50/1K scores, 14-day retention), Pro $249/mo, Enterprise custom.
Why it's great:
- CI/CD-integrated evals — blocks merges when quality drops
- AI-powered prompt optimization
- Impressive customer list: Notion, Stripe, Vercel, Airtable, Ramp
- Recently published EU compliance support (data plane separation, GDPR DPAs)
Why you might hesitate:
- $249/mo Pro is 12x what most indie founders are willing to pay
- Evaluation-focused — cost monitoring is secondary
- Starter tier's 14-day retention is limiting
- Stores prompts (even with EU data residency)
Best for: Well-funded teams that need CI/CD eval integration and have budget for enterprise tooling. Overkill for cost tracking alone.
3. Portkey (Best for Governance-Heavy Teams)
What it is: AI Gateway with governance layer. $18M+ raised ($15M Series A, February 2026).
Pricing: Gateway now FREE. Paid tiers from $49/mo for logging.
Why it's great:
- 1,600+ LLMs supported
- Virtual key management (secure key vault)
- Guardrails engine
- SOC 2, HIPAA, GDPR compliance
- Gateway recently made free for all teams
Why you might hesitate:
- Multiple G2/Gartner reviews cite bugs and weak documentation
- Cost tracking doesn't work for all models (user-reported)
- Still requires gateway routing (proxy lock-in, same as Helicone)
- Constant product changes — UI and features shift frequently
Best for: Enterprise teams that need governance, guardrails, and multi-LLM routing. If you liked Helicone's proxy approach but need more enterprise features.
4. AISpendGuard (Best for Privacy-First Cost Tracking)
What it is: AI cost monitoring with waste detection. Tags-only ingestion — never stores prompts. Bootstrapped.
Pricing: Free 50K events/mo. Pro €19/mo (500K events).
Why it's different from everything above:
- No proxy — passive SDK ingestion, zero latency impact, no lock-in
- No prompt storage — sends only tags (model, tokens, cost, custom labels). Ever.
- Waste detection — tells you which calls should use a cheaper model, with $/mo savings estimates
- EUR pricing — native for European teams, EU-hosted infrastructure (Frankfurt)
- Multi-framework — LangChain, LiteLLM, CrewAI, OpenClaw, OpenTelemetry
Why you might hesitate:
- No request logging — you can't debug prompts with AISpendGuard
- No evals, prompt management, or A/B testing
- Newer product — smaller community than Langfuse
- Bootstrapped — no VC war chest for rapid feature expansion
Best for: Indie founders and small teams whose main need is: "where is my AI money going and how do I spend less?" Privacy-conscious teams. GDPR-sensitive EU companies. Anyone who doesn't need prompt-level debugging.
5. Respan (formerly Keywords AI) (Rising Contender)
What it is: "Proactive AI observability" — automated eval agent that identifies regressions and recommends fixes. Rebranded from Keywords AI in March 2026 with $5M seed from Gradient Ventures (Google) + YC.
Pricing: Pro $0/mo (100K logs, 1K scores, 5 datasets). Team $249/mo. Enterprise custom.
Why it's great:
- Aggressive free tier ($0 Pro with 100K logs)
- Automated evaluation agent that recommends changes
- Auto-instruments LangChain, LlamaIndex, Vercel AI SDK, CrewAI, AutoGen
- Google-backed with growing feature set
- Publishing "Best Helicone Alternatives" content (actively capturing migration traffic)
Why you might hesitate:
- Just rebranded — product direction still stabilizing
- Stores full execution traces (messages, tool calls, routing, memory)
- Team tier jumps to $249/mo (same price as Braintrust Pro)
- Funding pressure may push further upmarket
Best for: Teams that want a generous free tier with automated quality monitoring. Good option if you don't mind prompt storage and want proactive recommendations.
How to Choose: Decision Tree
Did you use Helicone mainly for...
├── Debugging prompts / tracing chains?
│ ├── Budget > $200/mo? → Braintrust
│ └── Budget < $200/mo? → Langfuse (self-hosted or Cloud)
│
├── Cost tracking / attribution?
│ ├── Need GDPR compliance without storing prompts? → AISpendGuard
│ ├── Need governance + guardrails? → Portkey
│ └── Want free tier with automated evals? → Respan
│
└── Caching (Helicone's Redis cache)?
├── Anthropic? → Use built-in prompt caching (90% discount)
├── OpenAI? → Use built-in prompt caching (50% discount)
└── Need general caching? → Portkey or LiteLLM (self-hosted)
The Acquisition Risk Nobody Talks About
Helicone joins Langfuse (→ ClickHouse), Humanloop (→ Anthropic), and Promptfoo (→ OpenAI) on the acquisition list. The LLM observability market is consolidating fast.
When your monitoring tool gets acquired:
- Product goes to maintenance mode (Helicone) or gets absorbed (Humanloop)
- You scramble to migrate — while your monitoring is disrupted
- Your data is now owned by a different company with different privacy policies
This isn't an argument against any specific tool. It's an argument for:
- Open-source options (Langfuse, LiteLLM) where you own your deployment
- Tools that store less — if your monitoring tool only has your tags (not your prompts), migration is trivial and data exposure is minimal
- Avoiding gateway lock-in — proxy-based tools are hardest to migrate because they sit in your request path
Summary Table
| Tool | Pricing | Stores Prompts | Proxy Required | Waste Detection | Best For |
|---|---|---|---|---|---|
| Langfuse | Free – $2,499/mo | Yes | No (SDK) | No | Full observability, self-hosting |
| Braintrust | $0 – $249+/mo | Yes | Optional | No | Enterprise CI/CD evals |
| Portkey | Free gateway, $49+/mo logs | Yes | Yes | No | Governance, guardrails |
| AISpendGuard | Free – €19/mo | Never | No | Yes | Cost tracking, privacy |
| Respan | $0 – $249/mo | Yes | No (SDK) | Partial (quality-focused) | Free tier, automated evals |
What We'd Recommend
If you used Helicone for full observability: start with Langfuse. It's the most complete open-source alternative.
If you used Helicone mainly for cost tracking: try AISpendGuard. Free tier, 5-minute setup, no proxy to configure. You'll see where your money goes without exposing your prompts.
If you need enterprise governance: evaluate Portkey or Braintrust based on your budget and compliance requirements.
Whatever you choose, avoid building your own. Every "quick internal dashboard" becomes a maintenance burden within 3 months.
We built AISpendGuard to make cost tracking as simple as possible. Free tier, 50K events/mo, no credit card required. Try it →