Latitude pricing

The Tiers, as Published

Latitude prices its V2 agent observability platform in monthly credits, with unlimited seats on every plan.

Both paid-for features that security reviews ask about first sit on Enterprise: SAML SSO and role-based access control. Pro gets you the compliance reports, not the SSO.

What a Credit Is: Not Published

Latitude publishes the monthly credit allowances and the overage price, but not the conversion from product activity to credits. Without that conversion, a team cannot estimate how many sessions, traces or evaluations fit into Starter or Pro.

Before sizing a plan, get written answers to these:

• Which actions consume credits - ingested traces or spans, evaluation runs, LLM-judge calls, Agent Dispatch runs, or something else - and at what rate each?
• Does an LLM-as-judge evaluation cost credits on top of the traces it scores?
• What happens on Starter at 20K: does ingestion stop, or is the overflow dropped or queued?
• Is the $20 per 10K billed in whole blocks or pro rata?

What Pro Costs as Usage Grows

Assumptions for every row: Pro plan; 100K credits included in the $99; usage above that billed at $20 per 10K credits in whole 10K blocks; no taxes, discounts or Enterprise terms. The table prices credits, not sessions, because the conversion is not published.

Overage credits cost $0.002 each. The included credits should not be treated as costing $0.00099 each because the $99 plan also pays for retention, compliance reports and product access. At higher usage, compare the published overage total with an Enterprise quote. Seat count does not change the calculation.

The Self-Hosted Price: Zero, Plus Operations

The whole platform is MIT-licensed, and Latitude says every self-host option deploys the same images as its cloud. There is no licence fee and no credit meter. What you pay for instead is the stack: five application services (web, api, ingest, workers, workflows) on Postgres with pgvector, ClickHouse, Redis, Temporal and object storage, deployed with Docker Compose, Docker Swarm, a Helm chart or Railway. Add backups, upgrades, TLS and whoever is on call for it.

Teams that already operate Postgres, ClickHouse and Kubernetes may find self-hosting economical at higher volume. Others should include engineering time, upgrades, backups and incident response in the comparison. The operating detail is on Latitude self-hosting. On-premises deployment with Latitude support is an Enterprise feature.

How It Compares with Nearby Tools

Entry prices across the category: Failproof AI Team is $99 a month for 50,000 runs, 50,000 evals and 5 users, then $1.00 per 1k runs, with unlimited users at Scale for $599. Raindrop Pro is $299 a month plus $0.003 per event. Galileo, renamed Splunk Agent Observability in August 2026, lists Pro at $100 a month through a demo. Judgment Labs publishes no pricing.

Latitude combines unlimited seats with a fully MIT-licensed self-hosted platform. Its evaluations score completed sessions. Teams that also need automated failure analysis, alerts and a tested fix carried into agent behavior should compare the wider workflow in Failproof AI vs Latitude. Credits, runs, events and traces do not convert directly, so use your own traffic when comparing vendors. See agent observability pricing compared for the side-by-side.

When the Free Tier Is Enough

Starter gives a whole team unlimited seats and 20K credits a month with 30-day retention. For one agent in development, an internal tool, or a trial run against real traces, that is often all you need, and it costs nothing to find out how fast you burn credits before committing. If your data has to stay on your own infrastructure anyway, skip the cloud tiers and self-host.