Helicone sits in front of your model. We sit beside your app.
Helicone is an Apache-2.0 AI gateway and LLM observability platform: route through it, get per-request logs of cost, latency and errors. smolanalytics is one MIT Go binary that takes events from your app and computes product analytics and agent health together, with nothing in your request path.
smolanalytics vs Helicone, what is the difference?
They sit in different places. Helicone is an open-source (Apache 2.0) AI gateway and LLM observability platform: you send model calls through its proxy, which fronts 100-plus models, and it logs every request with cost, latency and errors, with an MCP server for querying those logs from an editor. Self-hosting it means Postgres, ClickHouse and MinIO alongside the dashboard and the Jawn API/proxy service. Note one fact before you pick it: Helicone announced on 3 March 2026 that it is joining Mintlify, and its own post says services stay live for the foreseeable future in maintenance mode, meaning security updates, new models, and bug and performance fixes keep shipping. Check the current state yourself before you build on it. smolanalytics is not a gateway and never touches your request path. It is a single MIT Go binary you send events to, and it does web and product analytics (funnels, retention, paths, cohorts, breakdowns, heatmaps, surveys, feature flags and A/B) plus computed agent observability off the same events: tool-call volume, error rate, latency p50/p90/p99, client split, an error taxonomy, and conversation health like turns, re-ask rate, abandonment and resolution. Every number is computed, never model-labelled, and a CI test pins /v1, MCP and the dashboard to the same answer. One MCP connection with an org token operates every project, 79 tools, answered by your own coding agent on your own model, so the AI is never metered by us. 14-day full trial with no card, Pro $49/mo for 1M events, Scale $149/mo for 10M, $8 per extra million, self-host free forever.
what each one actually is
“AI Gateway & LLM Observability Platform for AI Engineers”, a unified proxy across 100+ models, Apache 2.0.
github.com/Helicone/helicone, repo description and license
“Helicone's services will remain live for the foreseeable future in maintenance mode. This means security updates, new models, bug & performance fixes all keep shipping.”
helicone.ai, announcing it is joining Mintlify, 3 Mar 2026
smolanalytics is one MIT Go binary and one data file, actively developed. Product analytics and agent observability come out of the same events, with no proxy in front of your model calls.
what we do instead
Helicone vs smolanalytics
| Helicone | smolanalytics | |
|---|---|---|
| Where it sits | a gateway in front of model calls | an endpoint your app sends events to |
| What it records | per-request LLM logs, cost, latency | product events + computed agent health |
| Risk to your request path | in the path when you proxy | never in the path, send-and-forget |
| Model routing and key management | yes, that is the product | not offered, we are not a gateway |
| Funnels, retention, paths, cohorts | not its category | yes, same binary, same events |
| Agent rollups | dashboards over request logs | computed: p50/p90/p99, error taxonomy, re-ask, resolution |
| Self-host footprint | Postgres + ClickHouse + MinIO + 2 services | one static binary, one file |
| Ask from your editor | MCP over request logs | MCP over the same reports the dashboard renders, 79 tools |
| Dashboard vs API drift | separate read paths | CI pins /v1 == MCP == dashboard |
| License | Apache 2.0 | MIT |
when Helicone is the right tool
- You want a gateway, not an analytics store. Routing across many providers, one key surface, and logging every call at the proxy is Helicone's job and is not something we do at all. We never see your model calls, only the events your app chooses to send.
- You need the raw request and response of a single LLM call. Helicone keeps per-request bodies; we keep computed rollups over events, so we cannot show you what a specific call returned.
- You are already running it happily. It is Apache 2.0 and self-hostable, and its own post says services stay live in maintenance mode with security updates, new models and fixes still shipping. Read that post and the repo's current activity yourself, then decide, that is a call about your risk tolerance, not something we should make for you.
No proxy, one binary
# nothing in front of your model calls, nothing to route through docker run -p 8080:8080 -v $PWD/data:/data \ -e SMOLANALYTICS_DB=/data/smolanalytics.data \ ghcr.io/arjun0606/smolanalytics # dashboard + MCP at http://localhost:8080 # then, in your editor: "which tool has the worst p90 this week, and is its error rate rising?"
Your app keeps calling models directly. You send events, and get product analytics and computed agent health out of one process you own.
questions
Is smolanalytics a Helicone alternative?
For the observability half, yes. For the gateway half, no. Helicone is an Apache-2.0 AI gateway that proxies your model calls and logs each request with cost, latency and errors. smolanalytics is not a proxy: your app calls models directly and sends us events, and we compute agent observability from them (tool-call volume, error rate, latency p50/p90/p99, client split, error taxonomy, and conversation health like turns, re-ask rate, abandonment and resolution) alongside full product analytics. If you specifically want routing, key management and per-request logging at the proxy, that is Helicone's job, not ours.
Is Helicone still maintained?
Helicone announced on 3 March 2026 that it is joining Mintlify. Its own post says its services remain live for the foreseeable future in maintenance mode, which it defines as security updates, new models, and bug and performance fixes continuing to ship. The post does not spell out plans beyond that, and the project is Apache 2.0 on GitHub, so check the repo's current activity and their post directly rather than taking a competitor's word for it. We are not going to characterise their roadmap for them.
Do I have to route my LLM traffic through smolanalytics?
No, and you cannot. We are not a gateway and never sit in your request path, so we cannot slow down or break a model call. You send events from your app or agent, the same way you send any other analytics event, and the agent metrics are computed from those.
How do smolanalytics agent numbers stay trustworthy?
They are computed from your events, not labelled by a model, so nothing depends on an LLM's opinion. A CI test asserts the /v1 API, the MCP answer and the dashboard return the same number for the same question, so what your coding agent tells you in the editor cannot drift from the chart in the dashboard.
What does smolanalytics cost, and can I self-host it?
14-day full trial, no card. Pro is $49/mo for 1M events, Scale is $149/mo for 10M, and extra usage is $8 per million. Self-hosting is free forever under MIT: one Go binary, one data file, no external services. There is no free cloud tier.
Get the verdict, not another dashboard.
Drop one snippet, ask in your editor, own your data. 14-day full trial, no credit card.