Iris

How it works

Three simple ideas.

Iris is one workspace, with agents you can spin up in minutes β€” and every step they take is visible. You stay in charge.

One workspace

All your agents on one rail

Switch between your inbox helper, trip planner, and research buddy from a single sidebar. Same threads, same files, different specialists β€” no app-hopping.

Built in minutes

Make an agent without code

Pick a model, write a short instruction, choose what tools it can use. That’s it β€” your agent shows up next to the rest of the swarm and remembers the context you give it.

Always shows its work

No black box

Every reply lists what the agent looked at and what it ran. You can replay any step, follow a source, or ask why it picked that hotel β€” and you stay the one in charge.

Your first 60 seconds

  1. 1

    Open Iris

    Start with the built-in Iris 1.0 β€” ask it anything from a quick fact to a multi-step plan.

  2. 2

    Pick or build an agent

    Spin up a preset (Inbox helper, Trip planner, Research buddy…) or define your own in minutes.

  3. 3

    Watch the work

    Each reply shows the tools the agent used and any sources it pulled from. Replay any step.

  4. 4

    Stay sharp

    Approve, tweak, or hand off β€” agents wait for you on the decisions that actually matter.

Try it with one message.

No setup. No credit card. Open the workspace and ask the swarm something.