Agent platforms: your people and AI agents on the same work
We design and build agent platforms: shared workspaces where your people and many AI agents work on the same tasks. Everyone gets their own identity and their own view of the shared work. Agents keep working after you log off, approvals go straight to the one person who can give them, and every action is recorded. We build on Cloudflare (Workers, Durable Objects, Workflows) or on your own servers.
Why iRonin: we built the agent platform our own delivery runs on, and our work is merged into open-source agent tools other teams rely on.
What you can check
- Six of our pull requests are merged into Hermes WebUI, a web front end for the Hermes agent with over 19,000 GitHub stars. See them on GitHub →
- We maintain a public fork of herdr, a Rust terminal workspace for running coding agents, with our own releases. Releases →
- 25 of our pull requests are merged into open-source projects run by others, 19 of them into AI-agent tools. See them on GitHub →
- Rated 4.9 out of 5 on Clutch, from 33 reviews of our software and engineering work. Read the reviews on Clutch →
- Who it is for
- Companies putting AI agents to work across several people or teams, that need one shared place for the work, the approvals and the record of what happened.
What you get
- A separate identity for every person, device and agent — enrolled, granted and revoked.
- One shared workspace, scoped to what each person may see.
- Messaging and inboxes between agents and people, with delivery you can see.
- Approvals as their own channel: tied to one exact operation and one person who can decide it.
- Infrastructure keys kept on the platform, away from people's machines.
- A shared task board per workspace, safe under many writers at once.
Built with
Cloudflare Workers, Durable Objects and Workflows, or your own servers; services in Go and Rust; Claude, GPT, Gemini or open-weight models.
What we have built
The agent platform our own delivery runs on, on our own servers, now being extended to many users.
Questions
Why not one chat window per person?
Work that involves several people and several agents needs one shared state, so nothing depends on whose window is open.
Who approves what an agent does?
The person who owns that decision, for that exact operation. An approval is never an ordinary chat message.
Do agents hold our infrastructure keys?
No. We design it so keys stay on the platform and never reach an agent or a person's machine.