Pricing
Free to run. Nothing to buy yet.
Iris scores agent output for quality, safety, and cost, and today it runs entirely on your machine — MIT licensed, no evaluation limit, no account. Hosted and team features are under consideration, not under construction. If they ever ship, the evaluation will be the unit that matters, not the trace.
Open Source
Everything Iris does today, self-hosted
$0
MIT licensed — no limits, no account
- Unlimited evaluations — no metering, no quota
- All 20 built-in eval rules
- Custom Zod rules (unlimited)
- LLM-as-judge + citation verification (your API key, no proxy)
- Dashboard + playground
- stdio + HTTP transports
- Community support (GitHub Issues + Discord)
Runs entirely on your machine. Your traces never leave it.
Team
Shared history and hosted storage — planned
Not yet priced
nothing to buy today
- Everything in Open Source
- Managed hosting — no local database to run
- Shared team history across machines
- Trace comparison (side-by-side)
- Cost breakdown by agent + rule
- Agent-level dashboard filtering
Under consideration, not under construction. We will publish pricing when there is something to sell — and the open-source server will keep doing everything it does today.
Enterprise
For organizations with volume + compliance needs — planned
Not yet priced
nothing to buy today
- Everything in Team
- Single sign-on (SAML + OIDC)
- Custom eval rule authoring services
- Security review + procurement support
- On-premise / VPC deployment option
No compliance certifications have been started. If you need SOC 2 or similar today, we are not the right fit yet — the self-hosted server may still work for you, since it keeps all data on your own infrastructure.
FAQ
- What does it cost to run Iris today?
- Nothing. Iris is MIT licensed and runs on your own machine. There is no metering, no quota, no account, and no evaluation limit — the server does not count your usage, because there is nothing to count it for.
- Is LLM-as-judge a paid feature?
- No. LLM-as-judge and citation verification ship in the open-source server. They call Anthropic or OpenAI with your own API key and never route through us, so you pay your provider directly with no markup. A per-evaluation cost cap is enforced before each call, and the heuristic rules stay free and offline.
- Can I self-host?
- Self-hosting is the only way to run Iris right now — @iris-eval/mcp-server on npm or Docker, with the dashboard and playground in-process. Hosted options are described above as planned, not available.
- Will the open-source version get worse when a paid tier arrives?
- No feature that is free today will move behind a paywall. If a hosted tier ever ships, it will earn its price on hosting, shared team history and scale — not by removing something you already have.
- Do you have SOC 2 or other compliance certifications?
- No. None have been started, and we will say so plainly rather than imply otherwise. Because the server runs entirely on your infrastructure and sends no telemetry, many teams with compliance requirements can still use it — but that is your assessment to make, not a certification we hold.
- What is the waitlist actually for?
- It tells us whether hosted, team-shared storage is worth building. Joining costs nothing, commits you to nothing, and there is no price to be locked into yet.
Which one do I need?
The open-source server, because it is the only one that exists. It has no evaluation limit and no account, and it is what every feature listed above actually runs on. If shared team history or hosted storage would change your answer, tell us — that is how we decide whether to build it.
See the install docs →