At a glance
- Category
- Agent tools and integrations
- Tags
- Freemium, HIPAA, Open source, Self-hosted
Code
- Last push
- 15 October 2026
For developers
- Works with
- Shopify
What is Opik?
Opik is an AI observability and evaluation platform for AI teams that traces agent execution, tests outputs, and monitors production behavior. It logs user interactions, retrievals, tool calls, and model responses, then turns those traces into repeatable workflows with Trace & Debug Any Step, LLM-as-a-Judge Metrics, Test Suites & Assertions, Agent Playground, and Prompt Optimizer.
It integrates with Shopify and is used by Uber, Netflix, Etsy, and NatWest Group. Plans run Open Source Free, Free Cloud Free, Pro Cloud $19/month, and Enterprise custom.
What does Opik do?
Opik logs every step of an AI system, from user interactions and context retrieval to tool calls and model responses, then turns those traces into repeatable evaluation workflows. Teams can capture and visualize execution paths, define plain-text assertions or reference datasets, and use LLM-as-a-judge metrics to surface errors across development, testing, and production.
The result is a single workflow for tracing, debugging, and scoring agent behavior instead of stitching together separate tools. At scale, Opik supports 30+ metrics and is used by thousands of AI developers and teams, with 450 enterprise, startup, and academic teams and 150k+ developers cited across Comet's site.
It supports open-source self-hosting as well as hosted cloud plans, and the platform includes production monitoring, guardrails, token and cost tracking, and multi-media logging. Customers named on the site include Uber, Netflix, Etsy, Shopify, and NatWest Group.
Why use Opik?
- Opik combines tracing, evaluation, and production monitoring in one workflow, so teams can move from debugging to release decisions without switching systems.
- Its open-source codebase matches the hosted versions, which lets teams start free and keep the same workflow if they later self-host.
- The platform supports 30+ metrics and plain-text assertions, giving teams a structured way to compare agent versions and catch regressions.
- Production guardrails and audit logs help teams spot policy issues and create governance-ready records from the same traces they already collect.
- Enterprise plans add flexible deployments, SSO, service accounts, and compliance coverage for teams with stricter operational requirements.
Who is Opik for?
- ML engineers who need trace-level visibility into agent execution and failures.
- AI product teams who want repeatable tests before shipping model changes.
- Subject matter experts who review outputs and annotate problem traces.
- Platform teams who need production monitoring, guardrails, and cost tracking.
- Organizations with compliance needs that want self-hosting and flexible deployments.
What are Opik's key features?
Trace & Debug Any Step
Trace each agent step with AI observability and debugging views, so teams can inspect failures across complex workflows and fix issues faster.
LLM-as-a-Judge Metrics
Score outputs with LLM-as-a-Judge metrics and 30+ metrics helping teams compare quality signals without hand-reviewing every run.
Monitor Your Agents in Production
Track live agent behavior in production with tracing AI applications, giving teams visibility into regressions, latency, and unexpected tool use.
Test Suites & Assertions
Run evaluations with Test Suites and assertions to catch broken prompts or agent changes before release, using the same codebase as hosted versions.
Agent Playground
Experiment with agent behavior in the Agent Playground and compare runs before deployment, reducing trial-and-error in production systems.
Prompt Optimizer
Tune prompts with the Prompt Optimizer to improve agent performance, using evaluation results and tracing data to guide changes.
Shopify integration
Connect Shopify to trace and evaluate commerce workflows, so teams can inspect agent actions against real store operations and outputs.
What are Opik's use cases?
ML engineers trace failures
ML engineers use Opik to inspect agent runs step by step, using Trace & Debug Any Step to pinpoint where prompts, tools, or model outputs break. They can then compare traces and fix failure patterns before they reach users.
AI teams ship safer changes
AI product teams use Opik to validate model updates before release, using Test Suites & Assertions to run repeatable checks on expected outputs. They pair that with LLM-as-a-Judge Metrics to catch regressions that would otherwise slip into production.
Platform teams monitor production agents
Platform teams use Opik to watch live agent behavior, using Monitor Your Agents in Production to surface errors, drift, and cost spikes. They use the same traces to keep performance visible across deployments and respond faster when incidents happen.
Experts review traces and annotate
Subject matter experts use Opik to review problematic outputs, using Agent Playground to replay scenarios and annotate where responses go wrong. They can also use Prompt Optimizer to refine prompts based on those reviews and improve downstream quality.
How does Opik work?
- Connect your first application or agent and start capturing runs with Trace & Debug Any Step. Use the trace view to inspect each model call, tool action, and failure point.
- Add Test Suites & Assertions to define expected behavior for key workflows. Run them against new prompts or model versions before you ship changes.
- Score outputs with LLM-as-a-Judge Metrics to compare quality across traces and spot regressions. Review edge cases in Agent Playground when you need hands-on debugging.
- Turn on Monitor Your Agents in Production to watch live traffic, errors, and cost trends. Use the same dashboards to keep teams aligned on reliability and performance.
- Refine prompts with Prompt Optimizer, then rerun tests to confirm improvements. If you need broader rollout controls, use self-hosting and flexible deployments for compliance-sensitive environments.
How much does Opik cost?
Prices excluding tax, as Opik publishes them.
Open Source
Free
- Full AI observability & agent testing feature set
- True OSS: same codebase as the hosted versions
- Includes:
- Agent tracing & analysis
- Test Suites & assertions
- Agent Playground
Free Cloud
Free
- Up to 10 team members
- 25k spans per month
- 60-day data retention
- Includes:
- Agent tracing & analysis
- Test Suites & assertions
- Agent Playground
- Trial of Ollie, Comet's separate coding-harness product
Pro Cloud
$19 a month
- Up to 50 team members
- 100k spans per month
- 60-day data retention
- Includes everything in the Free plan plus:
- Customizable monthly span limits
- Customizable data retention periods
Enterprise
Custom
- Unlimited team members
- Custom usage plans
- Flexible deployments
- Service accounts and view-only users
- Single sign-on
- Dedicated support and SLAs
- SOC 2, ISO 27001, ISO 9001, HIPAA and GDPR compliance

Pricing page, captured
comet.com/site/pricing, checked
For AI agents
- MCP server
- YesChecked by us,
What we checked on Opik
What the price counts
Flat monthly rate of $19 for the Pro Cloud plan, with team-member and span ceilings rather than a per-seat charge, metered overage of $5 per additional 100k spans on top, and Enterprise sold only as a custom quote.
On the page: Pro Cloud plan card price cell, and Compare Plans table, Monthly Usage section, Additional Spans row
comet.com/site/pricingChecked
MCP server
MCP server at www.comet.com, behind OAuth sign-in
https://www.comet.com/opik/api/v1/mcp
comet.com/.well-known/oauth-protected-r…Checked by us,
Open source
Apache-2.0 licence in its public repository
Apache License Version 2.0
github.com/comet-ml/opik/blob/main/LICE…Checked by us,
Questions about Opik
- How much does Opik cost?
- Opik has a free plan. From comet.com/site/pricing, checked 5 November 2026.
- Does Opik have an API or an MCP server?
- Opik has an MCP server. We checked comet.com/.well-known/oauth-protected-resource ourselves on 15 November 2026.
- Is Opik open source, and can you host it yourself?
- Opik is open source. We checked github.com/comet-ml/opik/blob/main/LICENSE ourselves on 15 November 2026. You can host Opik yourself.
- What have we checked on Opik?
- 3 facts, under these headings: What it costs to start, how you build with it.
- How do we check Opik?
- We hold 3 facts on Opik, quoted from 4 pages. Facts last checked 15 November 2026. Price last checked 5 November 2026. Last change found 15 November 2026.
Sources and changes
Last full check
Sources on comet.com2
What changed1
- What it does
Repository
- Licence
- Apache-2.0
- Stars
- 22,631
- Last push
- 15 October 2026
Opik price history
The cheapest paid plan has cost $19 a month since 17 February 2026, on all 8 readings. Before that it was $39 a month.
| Price | Plan | First seen | Last seen | Readings | Source |
|---|---|---|---|---|---|
| $19 a month | 17 February 2026 | 20 November 2026 | 8 | Archived copy, 17 February 2026 | |
| $39 a month | 29 November 2024 | 17 November 2025 | 5 | Archived copy, 29 November 2024 |
The change we saw
- A paid plan $39 → $19 a month, seen between 17 November 2025 and 17 February 2026Archived copy, 17 November 2025Archived copy, 17 February 2026
From 13 readings of Opik's pricing page: 9 Internet Archive copies from 29 November 2024 and 4 of our own checks since 2 November 2026. A date is when we read the page, not when the price changed.
Similar in Agent tools and integrations
See all in Agent tools and integrations
- ERIUS PHONERented cloud Android phones, each in its own container, that keep an agent's apps and sign-ins between sessions.
- TinyFishWeb search, page fetching, cloud browsers and goal-driven web automation for AI agents, with search and fetch free.
- SmitheryPublic MCP server registry, now run by Arcade, listing direct vendor MCP URLs after its proxy shut down.
- UI DiscoveryA library of real websites, iOS apps and dashboards with each one's palette, fonts and original build prompt.
From our research
- Best sales agents for marketing teams, and what to checkApollo suits teams starting free, VoiceDrop suits voicemail campaigns, and PipeLime suits hands-off email and LinkedIn prospecting.
- Which browser automation API should you build an agent on?Skyvern, Browserbase, Browser Use, Playwright MCP, Hyperbrowser, and TinyFish compared on credential handling, session isolation, and audits, not speed.
- Clay pricing: every plan, its Actions, Data Credits and rowsWhat Clay's Free, Launch, Growth and Enterprise plans cost, and how many Actions, Data Credits and table rows each one includes.
Is Opik yours?
Free, and the page then carries the verified owner mark. You confirm it with an email at comet.com. Claim this listing
Report an error if a fact is wrong.