LLM observability and evals - verified
We read the pricing pages, docs, READMEs and license files of 20 LLM observability and evaluation tools on October 18, 2026: 13 hosted platforms, 2 open-source projects you run yourself, 3 add-ons to APM and analytics products and 2 eval frameworks. Every value carries a verbatim quote and its source. Prices stay in each vendor's own unit, so they are grouped by unit and never ranked across units.
Want the long read? See Langfuse vs Braintrust vs Helicone or the eval tools comparison. Routing model calls too? Compare LLM gateways.
The news in this category: acquisitions, renames, a shutdown, a pivot and a pricing change, each said only as far as the vendor's own page says it, with the quote and the link. Where the page gives no date, neither do we.
Langfuse says it has been part of ClickHouse since January 2026, and its license file now names ClickHouse, Inc. as the copyright holder.
“since January 2026 we're part of ClickHouse”
“Copyright (c) 2023-2026 ClickHouse, Inc.”
Promptfoo's README says it is now part of OpenAI and that it remains open source and MIT licensed. No date given.
“Promptfoo is now part of OpenAI.”
“Promptfoo remains open source and MIT licensed.”
Galileo's homepage says Galileo is now part of Cisco. No date given. On 2026-10-18 galileo.ai/pricing redirected to a Splunk pricing page for Splunk Agent Observability, which is where the Galileo price in this directory comes from.
“Galileo is now part of Cisco”
“Starting at just $100 per month per 1.2 million spans annually, billed annually”
Traceloop's pricing page carries a banner saying Traceloop is joining ServiceNow, written "Service now" on the page. No date given.
“Traceloop is joining Service now!”
Weave's pricing is now published on the CoreWeave Forge pricing page and its docs are published as CoreWeave Docs. No date given.
“CoreWeave Forge Pricing”
“Send OpenTelemetry Traces to Weave - CoreWeave Docs”
Helicone's pricing page carries a "Helicone Joins Mintlify" banner. We record the banner text only. No date given.
“Helicone Joins Mintlify”
Datadog's docs and pricing list present the product as Agent Observability, while the docs path still reads /llm_observability and the same page still links material titled Datadog LLM Observability. No date given.
“With Agent Observability, you can monitor, troubleshoot, and evaluate your LLM-powered applications, such as chatbots.”
“Agent Observability For first 100k LLM spans, per month, with 15-day retention by default For first 100k LLM spans, per month, with 15-day retention by default $160 $200 $240”
PostHog's docs are titled AI Observability, and the /docs/llm-analytics URL resolved to /docs/ai-observability when checked on 2026-10-18. No date given.
“AI Observability - Docs - PostHog”
Humanloop says its team is joining Anthropic and that, following the acquisition, the platform was sunset on September 8th, 2025, with all billing stopped since July 30th, 2025. Not included in this directory.
“We are thrilled to announce that the Humanloop team is joining Anthropic!”
“Following our acquisition, the Humanloop platform will be sunset on September 8th, 2025.”
“All billing has stopped since July 30th, 2025.”
Patronus AI's homepage now describes research into tasks and environments for reinforcement learning, and presents no evaluation or observability product or pricing. Not included in this directory. No date given.
“Our research connects a deep investigation of model failures with new methods for developing high-quality tasks and environments for reinforcement learning.”
Lunary's pricing page says new cloud workspaces require a paid subscription, and that this applies only to workspaces created after the change. No date given.
“New cloud workspaces require a paid subscription.”
“This payment requirement applies only to new cloud workspaces created after this change.”
Each card shows how the product charges in its own words, then its free tier, OpenTelemetry ingest, evals, self-hosting and license. Every value links the page it came from.
Managed tracing and evaluation platforms you send traces to. Several also publish the code to run yourself; those carry an open-source tag.
How it charges
Hobby: free - plan price
“Get started on the Hobby plan for free. No credit card required.”
How it charges
Developer: $0 per seat per month, then pay as you go - per seat per monthinferred
“$0 / seat per month then pay as you go”
How it charges
Starter: $0 per month - plan price per month
“Starter For everyone $0 / month”
How it charges
AX Free: free - plan price
“AX Free For builders Free”
How it charges
Free: $0 per month - plan price per monthinferred
“Designed for personal development of AI applications and models $0/mo”
How it charges
Hobby: free - plan price
“Hobby Free Kickstart your AI project.”
How it charges
Every plan includes $20/month of free usage (10M records); then $2 per million on Team and Growth - per million records (overage)
“Every plan includes $20 of free usage a month: 10 million spans, logs, or metrics.After that, $2 per million on Team and Growth.”
How it charges
Open Source: $0 - self-hosted
“Download, install, & run Opik your way $0 Free”
How it charges
Free: $0 per month, 1 GB data, no overage - plan price per month
“Free $0/ month 1 GB data no overage”
How it charges
Team: $20 per user per month - per user per monthinferred
“$20 / user / mo”
How it charges
Developer: free - plan price
“Free Start building Developer”
How it charges
Starting at $100 per month per 1.2 million spans annually, billed annually - per 1.2 million spans annually, billed annuallyvia redirect
“Starting at just $100 per month per 1.2 million spans annually, billed annually”
How it charges
Not statedNo public pricing page found on 2026-10-18
Projects whose primary form is code you run on your own infrastructure, with a license linked. Some add a hosted tier.
How it charges
Open source, self-hosted - self-hosted
“Phoenix is open-source and self-hosted.”
How it charges
Free Forever: $0 per month, up to 50K spans per month - plan price per monthinferred
“$0 / mo Up to 50K spans / month”
LLM and agent tracing inside a general observability or product analytics platform, billed in that platform's units.
How it charges
Price columns: billed annually / month-to-month / on-demand - billing column headers
“Product Billing Unit Billed Annually Billed Month-to-month Billed On-demand”
Every value, quote and source for Datadog Agent Observability
How it charges
Free: up to 20M events per month - events per month
“Free Event Volume: Up to 20M per month”
How it charges
AI Observability: 100K events free per month - events per month (free allowance)
“AI Observability 100K events”
Evaluation and testing libraries you run in code or CI, some with a hosted platform beside the open-source framework.
How it charges
Free: $0 forever - plan priceinferred
“Forever $0”
How it charges
Community: free forever - plan price
“Community Free Forever”
Every price line, in the vendor's own words, grouped by what it charges for. Rows are sorted by amount only inside one unit; a seat, a span, a Langfuse unit, a GB and a score are different things, so nothing is converted or ranked across units.
Flat plan and add-on prices, per month where the vendor says so. Free and custom plans are listed with them.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: plan price | ||
| Langfuse | Hobby: free | “Get started on the Hobby plan for free. No credit card required.” |
| Arize AX | AX Free: free | “AX Free For builders Free” |
| Helicone | Hobby: free | “Hobby Free Kickstart your AI project.” |
| Pydantic Logfire | Personal: free forever | “Personal Free, forever” |
| Opik | Free Cloud: $0 | “Free Cloud Perfect for individuals $0 Free plan” |
| HoneyHive | Developer: free | “Free Start building Developer” |
| Confident AI / DeepEval | Free: $0 foreverinferred | “Forever $0” |
| Promptfoo | Community: free forever | “Community Free Forever” |
| LangSmith | Enterprise: custom pricinginferred | “For teams with advanced hosting, security, and support needs. Custom pricing” |
| Braintrust | Enterprise: custom pricing | “Enterprise For teams at scale Custom pricing” |
| Arize AX | AX Enterprise: custom | “AX Enterprise For teams with scaled AI use cases Custom” |
| Helicone | Enterprise: contact us | “Enterprise Contact us Custom-built packages.” |
| Pydantic Logfire | Enterprise: custom | “Enterprise Custom” |
| Laminar | Enterprise: custom | “Enterprise Custom Custom limits On-premise” |
| Lunary | Enterprise: custom | “Enterprise Custom plans for larger teams with personalized needs.” |
| HoneyHive | Enterprise: custom pricing | “Custom Pricing Scale with confidence Enterprise” |
| Traceloop / OpenLLMetry | Enterprise: let's chat (>50K spans per month)inferred | “Let's chat >50K spans / month” |
| Confident AI / DeepEval | Enterprise: custom pricing | “Enterprise Custom pricing” |
| Promptfoo | Enterprise: custom | “Enterprise Custom For teams that need advanced features.” |
| Promptfoo | On-Premise: custom | “On-Premise Custom” |
| Unit: plan price per month | ||
| Braintrust | Starter: $0 per month | “Starter For everyone $0 / month” |
| Weights & Biases Weave | Free: $0 per monthinferred | “Designed for personal development of AI applications and models $0/mo” |
| Laminar | Free: $0 per month, 1 GB data, no overage | “Free $0/ month 1 GB data no overage” |
| Traceloop / OpenLLMetry | Free Forever: $0 per month, up to 50K spans per monthinferred | “$0 / mo Up to 50K spans / month” |
| Opik | Pro Cloud: $19 per monthinferred | “Expanded usage for teams $19 Per month” |
| Langfuse | Core: $29 per month | “Core For production projects. Longer data access and unlimited users. $29/ month” |
| Pydantic Logfire | Team: $49 per month | “Team $49/mo” |
| Arize AX | AX Pro: $50 per month | “AX Pro For AI-native teams $50per month” |
| Weights & Biases Weave | Pro: starts at $60 per monthinferred | “Starts at $60/month, billed monthly” |
| Helicone | Pro: $79 per month | “Pro POPULAR $79 per month” |
| Honeycomb | Pro: starting at $150 per monthinferred | “Starting at $150 per month” |
| Langfuse | Pro: $199 per month | “Pro For scaling projects. Unlimited history, high rate limits, all features. $199/ month” |
| Braintrust | Pro: $249 per month | “Pro For AI native teams $249 / month” |
| Pydantic Logfire | Growth: $249 per month | “Growth $249/mo” |
| Helicone | Team: $799 per month | “Team BEST VALUE $799 per month” |
| Langfuse | Enterprise: $2499 per month | “Enterprise For large scale teams. Enterprise-grade support and security. $2499/ month” |
| Unit: add-on price per month | ||
| Langfuse | Teams add-on: $300 per month | “Teams Add-on$300/mo” |
| Unit: plan price per month; per GB overage | ||
| Laminar | Starter: $30 per month, 3 GB included, then $2/GB | “Starter $30/ month 3 GB data included then $2 / GB” |
| Laminar | Pro: $150 per month, 10 GB included, then $1.50/GB | “Pro $150/ month 10 GB data included then $1.50 / GB” |
| Unit: per organization per month | ||
| Confident AI / DeepEval | Starter: $200 per month per organization | “Starter and Team are billed per organization per month at $200 and $2,000 respectively.” |
| Confident AI / DeepEval | Team: $2,000 per month per organization | “Team $2,000 / month” |
Prices charged for each seat or user.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: per seat per month | ||
| LangSmith | Developer: $0 per seat per month, then pay as you goinferred | “$0 / seat per month then pay as you go” |
| LangSmith | Plus: $39 per seat per month, then pay as you goinferred | “$39 / seat per month then pay as you go” |
| Unit: per extra seat | ||
| Pydantic Logfire | Team: 5 seats included, $25 per extra seatinferred | “5 seats included, $25/extra seat.” |
| Unit: per user per month | ||
| Lunary | Team: $20 per user per monthinferred | “$20 / user / mo” |
Usage priced per 100k. A Langfuse unit, an Opik span and a Datadog LLM span are not the same thing, so each keeps its own unit.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: per 100k units (overage) | ||
| Langfuse | 100k units per month included; additional units $8 per 100k, lower with volume | “100k units / month included, additional: $8/100k units. Lower with volume” |
| Unit: per 100k spans (overage) | ||
| Opik | Additional spans: $5 per 100k spans (Pro Cloud column of Monthly Usage table)inferred | “$5/100k spans” |
| Unit: per 100k spans retained | ||
| Opik | Additional span retention (60 to 400 days): $29 per 100k spans (Pro Cloud column)inferred | “$29/100k spans” |
| Unit: per first 100k LLM spans per month | ||
| Datadog Agent Observability | First 100k LLM spans per month (15-day retention default): $160 annually, $200 month-to-month, $240 on-demand | “Agent Observability For first 100k LLM spans, per month, with 15-day retention by default For first 100k LLM spans, per month, with 15-day retention by default $160 $200 $240” |
Datadog prices LLM spans per 10K, one row per trace retention period, in three billing columns.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: per 10K LLM spans per month | ||
| Datadog Agent Observability | 15-day retention: $3.50 / $4.20 / $5 per 10K LLM spans | “Per 10K LLM spans, per month, with 15-day trace retention and 3 month experiment retention $3.50 $4.20 $5” |
| Datadog Agent Observability | 30-day retention: $5 / $5.70 / $6.50 per 10K LLM spans | “Per 10K LLM spans, per month, with 30-day trace retention and 6 month experiment retention $5 $5.70 $6.50” |
| Datadog Agent Observability | 60-day retention: $6.50 / $7.20 / $8 per 10K LLM spans | “Per 10K LLM spans, per month, with 60-day trace retention and 9 month experiment retention $6.50 $7.20 $8” |
| Datadog Agent Observability | 90-day retention: $7.50 / $8.20 / $9 per 10K LLM spans | “Per 10K LLM spans, per month, with 90-day trace retention and 12 month experiment retention $7.50 $8.20 $9” |
Usage priced per 50k or 50M events, per million records, per 1.2 million spans, or in LangChain Standard Units. Different units; never converted.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: LSU per additional trace | ||
| LangSmith | Plus includes 10K traces total; additional traces 0.005 LSU eachinferred | “$39 / seat / mo · includes 10K traces total · 0.005 LSU / additional trace” |
| Unit: USD per LSU | ||
| LangSmith | 1 LSU = $1.00 | “LangChain Standard Units. $1.00 / LSU” |
| Unit: per million records (overage) | ||
| Pydantic Logfire | Every plan includes $20/month of free usage (10M records); then $2 per million on Team and Growth | “Every plan includes $20 of free usage a month: 10 million spans, logs, or metrics.After that, $2 per million on Team and Growth.” |
| Unit: per 50k events per month (overage) | ||
| Lunary | Team: 50k events included, then $10 per 50k per monthinferred | “Logs / Events 50k included then $10 / 50k / month” |
| Unit: per 1.2 million spans annually, billed annually | ||
| Galileo | Starting at $100 per month per 1.2 million spans annually, billed annuallyvia redirect | “Starting at just $100 per month per 1.2 million spans annually, billed annually” |
| Unit: per 50M events | ||
| Honeycomb | Pro: starting at $150 per 50M events (up to 750M) | “Pro: Starting at $150 / 50M events (up to 750M) and 250M time series data points (up to 3.75B)” |
Usage priced by data volume: processed, ingested or retained, per GB, per MB or per GB-month.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: per GB processed data | ||
| Braintrust | Pro: 5 GB processed data included, then $3/GBinferred | “5 GB processed data + $3/GB” |
| Braintrust | Starter: 1 GB processed data included, then $4/GBinferred | “1 GB processed data + $4/GB” |
| Unit: per GB per month retained | ||
| Braintrust | Pro: extended retention $0.50 per GB per monthinferred | “30-day retention + $0.50/GB/mo” |
| Unit: GB ingested per month | ||
| Weights & Biases Weave | Weave data ingestion: Free 1 GB/mo | “Weave data ingestion Volume discounts available with annual commitments Free 1 GB/mo” |
| Unit: per MB ingested (overage) | ||
| Weights & Biases Weave | Weave data ingestion: Pro 1.5 GB/mo, additional $0.10/MBas stated | “Pro 1.5 GB/mo Additional data ingestion $0.10/MB” |
| Weights & Biases Weave | Weave data ingestion: Enterprise additional $0.20/MBas stated | “Enterprise Additional data ingestion $0.20/MB” |
| Unit: per GB | ||
| Honeycomb | Telemetry pipeline: starting at $0.10 per GB | “Starting at $0.10 / GB” |
| Unit: per GB-month ingested or retained (overage) | ||
| Confident AI / DeepEval | Starter: 5 GB-months of trace spans, then $1 per GB-monthinferred | “5 GB-months of trace spans then $1 per GB-month ingested or retained” |
| Confident AI / DeepEval | Team: 75 GB-months of trace spans, then $1 per GB-monthinferred | “75 GB-months of trace spans then $1 per GB-month ingested or retained” |
Evaluation scores priced per thousand.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: per 1k scores | ||
| Braintrust | Pro: 50k scores included, then $1.50 per 1kinferred | “50k scores + $1.50/1k” |
| Braintrust | Starter: 10k scores included, then $2.50 per 1kinferred | “10k scores + $2.50/1k” |
Model usage inside the product: tokens analysed by a feature, or model credits.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: model credits | ||
| Braintrust | Starter: $10 model credits included, then token ratesinferred | “$10 credits + tok rates” |
| Unit: per million tokens analysed by Signals | ||
| Laminar | Signals: $0.05 per input Mtok, $0.01 per cached Mtok, $0.30 per output Mtok (Starter and Pro)inferred | “$0.05 / input Mtok $0.01 / cached Mtok $0.30 / output Mtok” |
What a plan includes before usage pricing, as each vendor states it.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: included usage | ||
| Arize AX | AX Pro: 50k spans per month, 10 GB ingestion per monthinferred | “Trace spans 50k spans per month - Ingestion volume 10 GB per month” |
| Unit: spans per month | ||
| Opik | Pro Cloud: 100k spans per month includedinferred | “Up to 50 team members - 100k spans per month” |
| Unit: included allowance | ||
| Galileo | Includes 500 million Luna tokens and 5 Luna fine-tuning runs per yearvia redirect | “Get 500 million Luna tokens and 5 Luna fine-tuning runs per year” |
| Unit: events per month | ||
| Honeycomb | Free: up to 20M events per month | “Free Event Volume: Up to 20M per month” |
| Unit: events per month (free allowance) | ||
| PostHog AI Observability | AI Observability: 100K events free per month | “AI Observability 100K events” |
| Unit: probes per month | ||
| Promptfoo | Community: red teaming up to 10k probes per monthinferred | “Red teaming (10k probes/month)” |
Free open-source and self-hosted editions, per-query add-ons and how billing is structured.
| Product | Price, in the vendor's words | Quote and source |
|---|---|---|
| Unit: self-hosted | ||
| Langfuse | Self-hosted: all core features free | “Self-host all core Langfuse features for free without any limitations.” |
| Opik | Open Source: $0 | “Download, install, & run Opik your way $0 Free” |
| Arize Phoenix | Open source, self-hosted | “Phoenix is open-source and self-hosted.” |
| Unit: usage-based | ||
| Helicone | Paid plans: usage-based pricing applies above included requests and storage | “10K free + usage-based” |
| Unit: per playground query (overage) | ||
| Lunary | AI Playground: 300 per month, then $0.05 per query | “AI Playground 300 / month then $0.05 per query” |
| Unit: open source | ||
| Traceloop / OpenLLMetry | OpenLLMetry SDK free (Apache-2.0) | “Yes! OpenLLMetry is Apache-2.0 licensed and you can use it for free” |
| Confident AI / DeepEval | DeepEval: open source | “DeepEval is a simple-to-use, open-source LLM evaluation framework, for evaluating large-language model systems.” |
| Unit: billing column headers | ||
| Datadog Agent Observability | Price columns: billed annually / month-to-month / on-demand | “Product Billing Unit Billed Annually Billed Month-to-month Billed On-demand” |
| Unit: billing note | ||
| PostHog AI Observability | Evaluations billed with AI observability | “Evaluations Billed with AI observability” |
Every product against every field, each cell linking its source. Listed on pricing page means the feature appears in a pricing table whose per-plan marks are icons, so no plan is claimed. Not stated is not the same as no. This is not a ranking.
OpenTelemetry ingest, tracing, evals, datasets, prompts, playground and cost tracking.
Self-hosting and license, retention, SOC 2 and HIPAA as listed, regions and SDKs.
Products whose own pages state or list each capability, in directory order. These are filters, not picks.
Accepts traces from your existing OpenTelemetry instrumentation.
Langfuse, LangSmith, Braintrust, Arize AX, Weights & Biases Weave, Pydantic Logfire, Opik, Laminar, HoneyHive, Arize Phoenix, Traceloop / OpenLLMetry, Datadog Agent Observability, Honeycomb
Scores production traffic, not only offline test sets.
Langfuse, LangSmith, Arize AX, Pydantic Logfire, Opik, HoneyHive, Galileo, Datadog Agent Observability, PostHog AI Observability, Confident AI / DeepEval, Promptfoo
Versions or serves prompts alongside traces.
Langfuse, LangSmith, Arize AX, Helicone, Opik, Lunary, HoneyHive, Arize Phoenix, Traceloop / OpenLLMetry, Confident AI / DeepEval
Code you can run yourself, license quoted and linked.
Langfuse, Weights & Biases Weave, Helicone, Pydantic Logfire, Opik, Laminar, AgentOps, Arize Phoenix, Traceloop / OpenLLMetry, PostHog AI Observability, Confident AI / DeepEval, Promptfoo
HIPAA appears on the vendor's own pages; check which plan carries it.
Langfuse, Braintrust, Arize AX, Weights & Biases Weave, Helicone, Pydantic Logfire, Opik, Laminar, HoneyHive, Confident AI / DeepEval
Values whose mapping is uncertain, flagged wherever they appear: usage prices placed by table position, a footnote of unclear scope, two rates shown exactly as stated, a page that disagrees with itself and a price read from a redirect. Separately, any plan name a quote does not carry is tagged inferred: the quote does not name this plan; the plan comes from the column or card the text sits in on the pricing page.
Weave data ingestion: Pro 1.5 GB/mo, additional $0.10/MB
Shown as stated: the page lists $0.10/MB for Pro and $0.20/MB for Enterprise additional ingestion. Not reconciled.
“Pro 1.5 GB/mo Additional data ingestion $0.10/MB”
Weave data ingestion: Enterprise additional $0.20/MB
Shown as stated: the page lists $0.20/MB for Enterprise and $0.10/MB for Pro additional ingestion. Not reconciled.
“Enterprise Additional data ingestion $0.20/MB”
Additional spans: $5 per 100k spans (Pro Cloud column of Monthly Usage table)
Assigned to Pro Cloud by its position in the Monthly Usage table; the quote itself does not name a plan.
“$5/100k spans”
Additional span retention (60 to 400 days): $29 per 100k spans (Pro Cloud column)
Assigned to Pro Cloud by its position in the Monthly Usage table; the quote itself does not name a plan.
“$29/100k spans”
Starting at $100 per month per 1.2 million spans annually, billed annually
galileo.ai/pricing redirected to this Splunk pricing page (Splunk Agent Observability) on 2026-10-18. The price is Splunk's, as that page states it.
“Starting at just $100 per month per 1.2 million spans annually, billed annually”
Includes 500 million Luna tokens and 5 Luna fine-tuning runs per year
From the Splunk pricing page that galileo.ai/pricing redirected to on 2026-10-18.
“Get 500 million Luna tokens and 5 Luna fine-tuning runs per year”
Up to 50K spans/month, up to 5 seats, 24 hours retention
The plan card says up to 5 seats; the FAQ on the same page says there is no seat limit. Both quotes are shown.
“Up to 50K spans / month Up to 5 Seats 24 Hours Data Retention”
FAQ: no seat limit, all features open
The FAQ says there is no seat limit; the plan card on the same page says up to 5 seats. Both quotes are shown.
“There is no seat limit, and all features are open.”
40k LLM spans are free (footnote)
A footnote on the pricing list. Its marker sits on the on-demand price for the first 100k LLM spans, and the page does not say which plans or billing options it covers.
“40k LLM spans are free.”
Products considered and left out, with the reason and the quotes recorded at the time.
Humanloop
Platform sunset after acquisition; no live product to evaluate
“We are thrilled to announce that the Humanloop team is joining Anthropic!”
“Following our acquisition, the Humanloop platform will be sunset on September 8th, 2025.”
“All billing has stopped since July 30th, 2025.”
Patronus AI
Site now describes reinforcement-learning training data and environments for AI labs; no evaluation or observability product or pricing is presented
“Our research connects a deep investigation of model failures with new methods for developing high-quality tasks and environments for reinforcement learning.”
Our longer reads on the products in this directory and on how to trace and evaluate agents.
Method
Read from each vendor's pricing pages, docs, READMEs and license files on , with every quote string-matched against the page text that day. Prices and plans change; check the linked source before you commit.
In the vendors' own words: Langfuse: "since January 2026 we're part of ClickHouse" Promptfoo: "Promptfoo is now part of OpenAI." Galileo: "Galileo is now part of Cisco" Traceloop / OpenLLMetry: "Traceloop is joining Service now!" Helicone: "Helicone Joins Mintlify" Humanloop: "We are thrilled to announce that the Humanloop team is joining Anthropic!" Only Langfuse (January 2026) and Humanloop (sunset September 8th, 2025) give a date on the pages we read.
Its license file says code outside the ee/ directories is available under the MIT Expat license, and it now names ClickHouse, Inc. as copyright holder. Langfuse's pricing page says: "Self-host all core Langfuse features for free without any limitations."
13 products state OpenTelemetry or OTLP ingest: Langfuse, LangSmith, Braintrust, Arize AX, Weights & Biases Weave, Pydantic Logfire, Opik, Laminar, HoneyHive, Arize Phoenix, Traceloop / OpenLLMetry, Datadog Agent Observability and Honeycomb. Not stated on the pages we read for Helicone, Lunary, Galileo, AgentOps, PostHog AI Observability, Confident AI / DeepEval and Promptfoo, which is not the same as unsupported.
12 products publish a license for code you can run: Langfuse (MIT Expat for code outside the ee/ directories), Weights & Biases Weave (Apache License 2.0), Helicone (Apache 2.0), Pydantic Logfire (MIT), Opik (Apache 2.0), Laminar (Apache License 2.0), AgentOps (MIT), Arize Phoenix (Elastic License 2.0), Traceloop / OpenLLMetry (OpenLLMetry: Apache 2.0), PostHog AI Observability (MIT Expat except the ee/ directory), Confident AI / DeepEval (DeepEval: Apache License 2.0) and Promptfoo (MIT). Check what each license covers: Langfuse and PostHog keep an ee/ directory under separate terms, Arize Phoenix uses the Elastic License 2.0, and Logfire's license covers the SDK while the platform is closed source.
In different units, so the prices do not rank: per seat (LangSmith, Lunary), per 100k units or spans (Langfuse, Opik), per 10K LLM spans (Datadog), per GB of processed or ingested data (Braintrust, Weave, Laminar, Confident AI), per 1k scores (Braintrust), per million records (Logfire) or a flat plan price. Every price line is listed in the vendor's own unit with its quote.
No. Every value is a first-party statement copied from the vendor's own pages, with the quote and the link. Nothing here measures trace volume, latency or eval quality, and prices in different units are never converted or compared.
Not stated means the pages we read did not say; it is not evidence the feature is missing. Listed on pricing page means the feature appears in a pricing table whose per-plan marks are icons, so we do not claim which plan has it. Inferred marks our reading: a plan name taken from the pricing table's layout, a usage price assigned to a plan by table position, or a status the page implies without labelling it.
Last verified from each vendor's own pricing pages, docs, READMEs and license files; every quote was string-matched against the page that day. This is not a ranking. Prices are shown in the vendor's own unit and change without notice, so check the linked source before you commit.
Snapshot of October 18, 2026: 20 products, 320 quotes, 12 with an open-source license, 2 considered and not included.
Newsletter
One email per week. New model releases, agent tooling changes and what they mean for what you build.
Routers and gateways that sit in front of your model calls: fees, BYOK, routing, caching and the logs they keep.
LangGraph, CrewAI, OpenAI Agents SDK and more: the code your traces come from.
Prompt caching, batch discounts and service tiers, provider by provider.
Our longer comparison of three tracing tools in this directory.