What is the difference between Iris and Latitude?

Iris is an MCP-native agent eval server that needs no SDK. Add the config block, restart your client, and every session lists Iris's tools on connect. Iris never intercepts: it runs when your agent calls one of its tools, when a host hook or `iris-eval ingest` hands it a trace, or when you POST one to its HTTP API. Latitude is an open-source, MIT-licensed platform for improving production AI agents: it captures agent traffic through a telemetry SDK or OpenTelemetry, groups repeated failures into

Is Iris better than Latitude for MCP agent evaluation?

Latitude offers things Iris does not: a free self-hostable platform with Sessions, Signals, Cost and Experiments views, LLM-as-judge and JavaScript evaluations that keep scoring live traffic, and a dispatch loop that hands failing traces to Claude Code or Cursor.

Iris vs Latitude — MCP-Native Agent Eval vs Open-Source Agent Observability Platform

TL;DR

Iris is an MCP server your agent discovers and uses on connect — one config block, no SDK, one SQLite file, every rule's precision and recall published at iris-eval.com/proof. Latitude — from its own pages, read 2026-09-21: Latitude is an open-source, MIT-licensed observability platform for production AI agents, operated by Latitude Data, S.L. in Barcelona. It ingests traces through a TypeScript or Python SDK or OpenTelemetry, groups failures into signals, scores sessions with LLM-as-judge, deterministic or JavaScript evaluations, tracks token cost, exposes an MCP server for coding agents, and can be self-hosted for free or run as Latitude Cloud. (source)

Iris grades what an agent did with its tools — the trace, the answer, the cost — not whether an MCP server honours its own contract; a server test harness answers that question, and Iris runs beside it. For the method, see the agent eval guide.

Feature comparison

Side by side.

Twelve features, the same twelve on every comparison. Every Latitude cell links the page it was read from and the date. The highlighted cells are Iris's own call on which side is stronger for a team running MCP agents — 3 to Iris, 2 to Latitude — not a measurement.

Feature	Iris	Latitude
Integration method	One block in the MCP config, no code — the agent discovers Iris and its tools on connect	TypeScript (@latitude-data/telemetry) or Python (latitude-telemetry) SDK; OpenTelemetry exporter for other languagesLatitude's page · read 2026-09-21
Self-hosting	One process, one SQLite file; Docker image with a health check	Free self-hosting (Docker Compose or Helm): five services + Postgres/pgvector, ClickHouse, Redis, Temporal, object storeLatitude's page · read 2026-09-21
Where it runs	Nothing in the agent's process — Iris is a separate server the agent calls	In-process telemetry SDK in your app, or your existing OpenTelemetry pipeline exporting to Latitude; no proxy describedLatitude's page · read 2026-09-21
Evaluation	21 built-in deterministic rules and 9 custom-rule types, in-process; 6 judge templates on a key you supply; every rule's precision and recall published	LLM-as-judge, deterministic conditions or custom JavaScript score completed sessions; human annotations check alignmentLatitude's page · read 2026-09-21
Cost tracking	Per-trace USD cost and tokens; a cost spike judged against the agent's own history	Input/output/total tokens and estimated cost per trace and session; Cost dashboard with per-model spend, cache adviceLatitude's page · read 2026-09-21
MCP support	Protocol-native — Iris is an MCP server with 12 tools; OTLP traces in	Remote OAuth MCP server (api.latitude.so/v1/mcp) with tools generated from the Latitude API, for coding agentsLatitude's page · read 2026-09-21
License	MIT, the whole package	MIT on the default branch (development); the older main branch still carries an LGPL-3.0 LICENSE and READMELatitude's page · read 2026-09-21
Ownership	Independent and founder-led	Operated by Latitude Data, S.L., Barcelona, Spain; funding and investors are not stated on its pagesLatitude's page · read 2026-09-21
Dashboard	A local dashboard on its own port: traces, moments, regressions, five views	Web UI with Traces, Sessions, Signals, Evaluations, Cost, Users, Tools, Memory, Search, Behaviors and Experiments viewsLatitude's page · read 2026-09-21
Framework support	Any MCP client (1 verified, 9 claimed — see /clients); OTLP/HTTP from anything else	OpenAI, Anthropic, Bedrock, Vercel AI SDK, LangChain and more, plus any OpenTelemetry-compatible applicationLatitude's page · read 2026-09-21
Prompt management	Not included	Not stated in the vendor's documentation as of 2026-09-21Latitude's page · read 2026-09-21
Enterprise and compliance	Self-hosted. Nothing leaves your machine unless you set IRIS_OTEL_ENDPOINT, which exports traces to the collector you name, or enable the LLM judge with your own key. No compliance certification is claimed before it is held	SOC 2 Type II; ISO 27001 reports; GDPR; SAML SSO, RBAC, custom on-prem or cloud deployment on the Enterprise planLatitude's page · read 2026-09-21

Decision guide

Which one fits your stack?

When to choose Iris

You are building with MCP-compatible agents and want the integration to be one config block
You want the evaluation to be deterministic and local — no model call, nothing leaving the machine
You want to read what each rule is worth before you trust it: every rule's precision and recall is published
You want self-hosting to be one process and one file
You want a fully permissive MIT license on the whole package

When to choose Latitude

You need to self-host the entire platform for free on Docker Compose or Kubernetes, with your data staying in your own infrastructure. source
You need recurring production failures auto-grouped into tracked signals with status, size and trend, Sentry-style. source
You need a fix loop that dispatches Claude Code or Cursor with the failing traces and replays the fix against those traces. source
You need a Cost dashboard with per-model economics and cache recommendations computed from your own traffic. source
You need OpenTelemetry-native ingestion from any language, not only Python or TypeScript. source

Sources

Every Latitude statement on this page was read from one of these pages on the date shown. The file behind this page is website/src/lib/compare/latitude.json.

Latitude homepage 2026-09-21
Latitude pricing 2026-09-21
Latitude about 2026-09-21
Latitude privacy policy (legal entity) 2026-09-21
Latitude docs index (llms.txt) 2026-09-21
Latitude docs introduction 2026-09-21
Latitude MCP server docs 2026-09-21
Latitude start tracing 2026-09-21
Latitude deployment overview 2026-09-21
Latitude evaluations overview 2026-09-21
Latitude evaluation detection methods 2026-09-21
Latitude evaluation alignment 2026-09-21
Latitude token and cost tracking 2026-09-21
Latitude Cost view 2026-09-21
Latitude SOC 2 page 2026-09-21
latitude-dev/latitude-llm repository 2026-09-21
latitude-llm repository metadata (GitHub API) 2026-09-21
latitude-llm LICENSE (development branch, MIT) 2026-09-21
latitude-llm README (development branch) 2026-09-21
latitude-llm LICENSE (main branch, LGPL-3.0, stale) 2026-09-21
latitude-llm README (main branch, stale) 2026-09-21

Ready to see what your agents are doing?

Add Iris to your MCP config. First trace in 60 seconds. No SDK, no signup, no infrastructure.

Install Iris Every comparison