The main Braintrust alternatives for agent analytics are Langfuse, Arize Phoenix, LangSmith, Galileo (now Splunk Agent Observability), Opik and Kubit. Langfuse, Phoenix and Opik are open source and self-hostable. LangSmith and Galileo are managed platforms. Kubit queries your traces where they already are, labels and scores every span, and visually groups runs by execution path, with no additional LLM judge cost and no sampling.

Key facts (as of 16 Nov 2026)

  • Braintrust prices on processed data and scores: Starter is free, Pro is $249 a month, and self-hosting is Enterprise only (pricing)

  • Open source options: Langfuse (MIT outside its enterprise folders), Arize Phoenix (Elastic License 2.0) and Opik (Apache 2.0)

  • Cheapest paid cloud entry: Opik Pro at $19 a month, then Langfuse Core at $29 (Comet, Langfuse)

  • Kubit is in early access and works alongside Braintrust through its agent plugin (docs)

What Braintrust does well

Braintrust is strong at evaluation and at making traces searchable for a whole team: users are unlimited on every plan. Its Topics feature has an LLM summarize each trace through three lenses (task, sentiment, issues), embeds the summaries and clusters them, so you can see what conversations were about and where users were unhappy. Patterns, in public preview, has its Loop agent read a varied sample of traces and save recurring problems with evidence. Braintrust Topics groups runs by a summary of what they were about. It doesn't group them by the path they took.

Two ways to group agent runs: Braintrust Topics clusters an LLM summary of each trace, while Kubit Flows groups every run by execution path in one Sankey.

Braintrust alternatives compared

Braintrust alternatives compared: open source, self-hosting, pricing and free tier for seven tools, plus how Braintrust Topics and Kubit Flows group the same runs.

Tool

Open source

Self-hosting

How it prices

Free tier

Braintrust

No

Enterprise plan

Processed data (GB) and scores

Starter: 1 GB, 10k scores

Langfuse

Yes, MIT core

Free

Units (traces, observations, scores)

Hobby: 50k units

Arize Phoenix

Yes, ELv2

Free

AX cloud: spans and GB

AX Free: 25k spans

LangSmith

No

Enterprise plan

Seats and traces

Developer: 5k traces

Galileo (Splunk)

No

Not stated

Not published

Free edition

Opik

Yes, Apache 2.0

Free

Spans

Free Cloud: 25k spans

Kubit

No

Reads your warehouse in place

Early access

Early access

Prices from each vendor's pricing page, 16 Nov 2026.

Which Braintrust alternatives are open source and self-hosted?

Three. Langfuse is self-hostable for free with Docker Compose or Kubernetes, and its core is MIT licensed. Arize Phoenix is open source and local-first, under the Elastic License 2.0. Opik runs the same codebase as its hosted version under Apache 2.0. Braintrust and LangSmith both offer self-hosted deployments on their Enterprise plans.

How do Langfuse, Phoenix, LangSmith, Galileo and Opik compare?

  • Langfuse is the most widely adopted open source option, with tracing, evals and prompt management in one product and more than 37,000 GitHub stars.

  • Arize Phoenix suits teams who want to start locally and later move to Arize AX, whose free tier includes 25k spans.

  • LangSmith fits teams already building on LangChain or LangGraph, priced per seat with a free developer tier.

  • Galileo, now Splunk Agent Observability, combines evaluation, cost tracking and runtime guardrails that can block risky actions before they run.

  • Opik has the lowest paid cloud entry at $19 a month and doesn't charge per seat.

How is Kubit different?

Kubit is Agent Analytics: agent observability with behavior analytics. Polaris, Kubit's in-house small language model, labels and scores every span in under 100ms, with no extra LLM cost. It is up to 98% lower cost and ~15x faster than a frontier LLM judge, with comparable accuracy on public agent benchmarks. Polaris marks Missteps (every call returned OK and the agent still did the wrong thing) and Exceptions (execution failed). Flows visually groups every run by execution path, meaning which tools ran in what order, and draws all paths in one Sankey. Runs then join to sessions, users and outcomes in funnels, retention and cohorts. If your traces are already in Snowflake, Databricks, BigQuery or ClickHouse, Kubit queries them there. See How Kubit works.

Can I keep Braintrust and add Kubit?

Yes. The Kubit agent plugin for Claude Code or Cursor finds your existing tracing setup and adds Kubit's span processor next to your Braintrust processor, so Braintrust keeps receiving every span. It shows a diff before every file it changes. See Integrate with the agent plugin.

How to choose

Pick an open source tool if self-hosting for free matters most. Pick LangSmith if your agents run on LangChain. Pick Galileo if runtime guardrails are the priority. Add Kubit when you need to see which execution paths fail, including the runs that return OK, and what users did after them.

Kubit's Agent Analytics is in early access. Join the waitlist or book a time with Alex.

Read next: What is Kubit? · Why we group agent runs by execution path · Why we label every span instead of a sample

FAQ

What are the best Braintrust alternatives in 2026?

The main alternatives are Langfuse, Arize Phoenix, LangSmith, Galileo (now Splunk Agent Observability), Opik and Kubit. Langfuse, Phoenix and Opik are open source and self-hostable. LangSmith and Galileo are managed platforms. Kubit queries traces in your warehouse and visually groups runs by execution path.

Is there an open source alternative to Braintrust?

Yes. Langfuse (MIT core), Arize Phoenix (Elastic License 2.0) and Opik (Apache 2.0) are open source and can be self-hosted for free. Braintrust offers self-hosting only on its Enterprise plan.

How does Braintrust group traces?

Braintrust Topics has an LLM summarize each trace through facets such as task, sentiment and issues, embeds the summaries and clusters them into topics. Patterns, in public preview, reads a sample of traces and saves recurring problems with evidence.

Can I use Kubit with Braintrust?

Yes. The Kubit agent plugin adds Kubit's span processor next to your Braintrust processor, and Braintrust keeps receiving every span.

How much does Braintrust cost?

As of 16 Nov 2026, Braintrust Starter is free with 1 GB of processed data and 10k scores, Pro is $249 a month, and Enterprise is custom. Users are unlimited on every plan.

Table of contents

Running agents in production?

The first cohort is closed. Join the waitlist for the next one. Your traces stay in your warehouse.

Summarize this post

Share