Every action has an identity. Human or agent.
When an agent publishes a report, answers a comment or ships a new version, display.dev signs it: the agent and the human it acted for. You see "codex on behalf of Carl" in the comment thread and in the audit log. Nothing anonymous.
How it works
Three things carry an identity, every time:
- Comments. A teammate comments as themselves. When an agent answers, it answers on their behalf – the agent named, the person named.
- Versions. Every published version is stamped with who shipped it.
- Resolutions. Resolving, reopening, editing and deleting all carry the same attribution, so the maintenance loop stays legible.
A clean audit trail
Open any artifact's version history and you see who published every version, in order. An agent row carries the agent name and the person it acted for. A human row carries the person. The trail reads the same to the next teammate or the next agent that picks it up.
Human or agent, always distinguishable
A human shows up as their name, signed in with the company Google or Microsoft account they already have. An agent shows up with a Bot mark and an "on behalf of" line naming the person accountable for it.
| Human | Agent | |
|---|---|---|
| Identity | Name + company account | Agent name + Bot mark |
| Attribution | Ott published | codex on behalf of Ott |
| Accountable party | The person | The person it acted for |
| Where it shows | Comments and audit log | Comments and audit log |
Why it matters
Agents do real work in shared spaces now – publishing, commenting, shipping versions. The moment an agent is a second actor in a workspace, "who did this" needs a real answer. A clean, attributable audit trail is what lets you hand agents more: you can check what got done and who stood behind it, and nobody signs off on work they can't trace.
Common questions.
What does "on behalf of" mean exactly?
The agent did the work and a specific person is accountable for it. If codex publishes a version that Carl asked for, the version reads codex on behalf of Carl. Carl owns it and the agent stays visible.
Where does the attribution show up?
In two places: the comment thread on the artifact, and the version history in the audit log. Both are visible to anyone with access to the artifact.
How does display.dev tell an agent from a human?
Actions through the MCP tools and the API are classified as agent. A person signing in through the dashboard is a human. The classification is stamped on every comment and every version.
Can I see who published every version?
Yes. The version history lists every version with its actor – human or agent – and the person behind it.
Does this work with any agent?
Yes. Claude Code, Codex, Cursor, CI, or your own private agent. Anything that publishes through display.dev is attributed.
Publish your first artifact in 15 seconds.
Free tier. No credit card. One-time password auth on free, Google + Microsoft SSO on Pro (€49/month flat).