Braintrust scores your agent's outputs. We count what your product and your agent did.
Different jobs, and we will say so plainly: Braintrust is the deeper tool for evals, datasets, tracing and prompt experiments, and it is built for enterprise AI teams. smolanalytics is one MIT Go binary that does product analytics and computed agent observability from the same events, with no scoring pipeline, no control plane, and no metered AI.
smolanalytics vs Braintrust, which one do I want?
It depends on what you are trying to know, and some teams will want both. Braintrust (braintrust.dev, the AI platform, not the freelancer network) is a closed-source enterprise platform for shipping quality AI agents: trace prompts, responses and tool calls, run evals scored by LLMs, code or humans, manage datasets, experiment in a playground, and surface production patterns to turn into new evals. Notion, Vercel, Replit and Dropbox are among its named customers, and it raised an $80M Series B led by ICONIQ in February 2026. It has a real free tier (1 GB of processed data, 10k scores, 14-day retention), Pro is $249/mo, and self-hosting is hybrid: on Enterprise you run the data plane, the Braintrust API plus PostgreSQL, Redis, object storage and the Brainstore query engine, on your own infrastructure while the web UI, auth and user management stay on Braintrust's hosted control plane. If you need evals, that is Braintrust, and it is better at it than we will ever be. smolanalytics is a different shape: a single MIT Go binary that does web and product analytics (funnels, retention, paths, cohorts, breakdowns, heatmaps, surveys, feature flags and A/B) and computed agent observability off the same event stream, 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. Those numbers are computed from your events, never scored or labelled by a model. You ask it questions in plain English from your editor: 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. Pricing is a 14-day full trial with no card, Pro $19/mo for 2M events, Scale $49/mo for 5M, $6 per million, self-host free forever.
what each one actually is
“Ship quality agents at scale.” “Surface patterns in production, turn them into evals, and improve quality with every release.”
braintrust.dev, homepage hero
Self-hosting Braintrust is hybrid: you run the data plane, the Braintrust API, PostgreSQL, Redis, object storage and the Brainstore query engine, on your infrastructure, while the web UI, authentication and user management stay on Braintrust's hosted control plane.
braintrust.dev/docs/guides/self-hosting, deployment model
smolanalytics is one MIT Go binary and one data file, fully yours. Product analytics and agent observability come out of the same events: no scoring pipeline, no query cluster, no hosted control plane.
what we do instead
Braintrust vs smolanalytics
| Braintrust | smolanalytics | |
|---|---|---|
| The job it is built for | evals, tracing, prompt experiments | product analytics + agent health |
| Evals, datasets, scorers | core product: LLM, code or human | not offered |
| Prompt playground | yes, unlimited on every plan | not offered |
| Trace-level LLM debugging | deep: prompts, responses, tool calls | no, we compute rollups |
| Funnels, retention, paths, cohorts | not its category | yes, same binary, same events |
| Agent numbers | traces plus scored evals | computed: p50/p90/p99, error taxonomy, re-ask, resolution |
| Open source | platform closed, SDKs Apache-2.0 | MIT, all of it |
| Self-host footprint | hybrid: API + PG + Redis + S3 + Brainstore, control plane stays theirs | one static binary, one file, all yours |
| Ask from your editor | web UI, SDKs and API | MCP, one org token routes every project, 79 tools |
| Dashboard vs API drift | separate read paths | CI pins /v1 == MCP == dashboard |
| Entry price | free tier, then Pro $249/mo | free tier, then Pro $19/mo, self-host free |
when to use Braintrust (often alongside us)
- You run evals. Scoring outputs with LLMs, code or humans, managing datasets, running experiments and iterating in a playground is Braintrust's whole product, and its Starter plan gives you unlimited projects, datasets, playgrounds and experiments for free. We ship none of that and are not trying to.
- You need to see inside a single run. Braintrust traces prompts, responses and tool calls in real time, and its Loop agent and Topics surface patterns in production to turn into new evals. We compute rollups over your events and do not store per-call LLM traces, so if the question is “what exactly did the model say in that run”, that is Braintrust, not us.
- You are an enterprise AI team. Braintrust names Notion, Vercel, Replit and Dropbox among its customers, raised an $80M Series B led by ICONIQ in February 2026 to, in its own words, “become the observability layer for production AI”, and offers hybrid deployment where sensitive data stays in a data plane on your infrastructure. Running both is a sane setup: Braintrust for quality, smolanalytics for the product usage and the computed agent health beside it.
Add the other half in 30 seconds
# no Brainstore, no Postgres, no Redis, no hosted control plane 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: "did users who hit a tool error this week come back, and which tool was it?"
Keep Braintrust for evals and quality. Put this next to it for funnels, retention and the computed agent numbers, on your own box, answered by your own model.
questions
Is smolanalytics a Braintrust alternative?
Only partly, and we would rather be straight about it. Braintrust is an enterprise AI evals and observability platform: trace every call, score outputs with LLM, code or human evaluators, manage datasets, and experiment in a playground. It is deeper than us at all of that. smolanalytics is one MIT Go binary that gives you web and product analytics (funnels, retention, paths, cohorts, heatmaps, surveys, flags and A/B) plus computed agent observability from 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. If evals are what you need, use Braintrust. If you want product usage and agent health in one place with no extra infrastructure, use us. Some teams run both.
Can smolanalytics replace Braintrust for evals?
No. We do not score outputs, we do not manage datasets or experiments, we do not ship a prompt playground, and we do not store per-call LLM traces. We are not planning to. Our agent observability is computed rollups over the events your app already sends, which answers “is the agent healthy and where is it failing”, not “is this answer good”. For that second question, Braintrust.
Is Braintrust open source? Can I self-host it?
The platform is closed source; Braintrust open-sources its SDKs (Apache-2.0) and tools like autoevals and its proxy, but not the product itself. Self-hosting exists and is hybrid: on Enterprise you deploy the data plane, the Braintrust API plus PostgreSQL, Redis, object storage and the Brainstore query engine, on your own AWS, GCP or Azure via Terraform, while the web UI, authentication and user management remain on Braintrust's hosted control plane. Their docs state the control plane does not store or process your sensitive data. Self-hosting smolanalytics is one MIT binary and one file, with nothing that phones home and no plan requirement.
How are smolanalytics agent numbers different from Braintrust evals?
Braintrust measures quality: an LLM, a script or a human scores each output against your criteria, which is genuinely useful and genuinely a judgment. Our numbers are counted, not judged: tool-call counts, error rate, latency percentiles, client split, the error taxonomy, turns, re-ask rate, abandonment and resolution are all derived deterministically from your event stream, so nothing is a model's opinion and nothing changes when a model changes.
Is this the Braintrust freelancer network?
No. This page compares braintrust.dev, the AI evals and observability platform used by teams like Notion and Vercel, which raised an $80M Series B led by ICONIQ in February 2026. It is unrelated to usebraintrust.com, the talent marketplace of the same name.
What does smolanalytics cost, and can I self-host it?
There is a free tier (1M events/mo, no card) and a 14-day full trial of Pro on top of it. Pro is $19/mo for 2M events, Scale is $49/mo for 5M, and extra usage is $6 per million. Self-hosting is free forever under MIT: one Go binary, one data file, no external services. There is no free cloud tier; Braintrust's Starter plan is free with 1 GB of processed data, 10k scores and 14-day retention, so if a free hosted tier is the deciding factor, they have one and we do not.
Get the verdict, not another dashboard.
Drop one snippet, ask in your editor, own your data. 14-day full trial, no credit card.