Coralogix AI Observability
AI observability with per-agent evaluators, cost tracking, and session tracing
- Tool type
- AI observability software with evaluators and dashboards
- Made for
- Engineering and operations teams running AI agents
- Runs on
- Coralogix web dashboard
- Accessed through the Coralogix web UI; no desktop, mobile or self-hosted runtime is stated.
What it does
Coralogix AI Observability, part of its AI Center, takes in the prompts, responses, sessions, token usage and costs produced by AI agents. It runs evaluators that score each message in real time for issues such as hallucinations, PII leaks, relevance and toxicity, and it presents the results in dashboards and a session explorer. Users can trace a session from start to finish, see which evaluators flagged which messages, and break spend down per message, session and agent.
Teams use it to find errors or sudden quality drops in individual agents, to debug or audit how users interact with agents, and to catch inefficient prompts, token overuse and unusual usage spikes before they affect budget. Evaluators can be out-of-the-box or custom and are assigned per agent. Dashboards cover evaluator results, token usage, errors, latency and user feedback, and results can be shared with a team.
It suits engineering and operations teams running AI agents in production who already use or plan to use Coralogix's observability platform. Access starts with a free 14-day trial with all features and no credit card. AI evaluation is billed on tokens: $1.50 per 1M tokens, and tokens are measured monthly and apply only to AI evaluation workloads. Unused tokens expire at the end of the subscription term and cannot be carried forward or refunded.
What it can do
- Real-time evaluation of every AI message using out-of-the-box or custom evaluators such as hallucinations, PII leaks, relevance and toxicity
- Dashboards combining evaluator results, token usage, errors, latency and user feedback
- Session explorer to trace a user session from start to finish with evaluator flags per message
- Cost tracking broken down per message, session and agent to surface overspend and suspicious usage
- Index-free remote querying and infinite retention via the customer's S3 bucket
Before you choose
- AI evaluation consumes tokens measured monthly; processing one item through three evaluators consumes three times the tokens.
- Unused Units or Tokens expire at the end of the subscription term and cannot be refunded or carried forward.
- Logs, traces and metrics are billed in units on a daily quota, separate from the monthly AI token meter.
Listed under
Get started
Free trial with all features and no credit card required.
- Sign up for the free trial at Coralogix (no credit card needed) to get 14 days of full feature access with a quota of 8 units.
- Get started with the Coralogix platform using the getting-started documentation.
- Assign out-of-the-box or custom evaluators to each AI agent in the AI Center evaluation engine.
Plans and prices
Prices in USDFree 14-day trial with a quota of 8 units and no credit card. AI Center Tokens are measured monthly and apply only to AI evaluation workloads. Unused Units or Tokens expire at the end of the term and have no monetary value. Confirm current rates on the pricing page.
- Real-time evaluation of AI messages
- Tokens measured monthly
14-day free trial.
Source: coralogix.com/pricing, checked 30 Oct 2026. Prices can change; check before you buy.
Sources for this overview
Product pages, documentation, and pricing checked 30 Oct 2026. Each source supports part of the overview; availability and prices can change.
- AI observability - Coralogix ↗coralogix.com
- Pricing - Coralogix ↗coralogix.com
Ratings and reviews
No ratings for Coralogix AI Observability yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- coralogix.com: product, pricing and policy pages
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- knowledge_base
- manual_review
Last checked 30 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.