The product

My Agents

The dashboard where all your agents live. Three tabs to separate what you've created, what you've installed from the marketplace, and your multi-agent workflows.

The 3 tabs

Created

The agents you designed yourself (from the builder or a template). Without fork lineage — this is your original work. If you publish one of these agents on the marketplace, a sage chip published · /slug appears on its card, clickable through to the public page.

Installed

The agents that are forks of other published agents. They keep a lineage link to their parent — visible on each card via a butter forkedchip. The lineage lets you find the original in 1 click even if you've heavily modified it.

Workflows

Multi-agent chains that you've saved. Each workflow chains N agents in order: the output of agent 1 becomes the input of agent 2, and so on. See Webhooks & workflows for the details.

On each card

  • A monogram with the initial of the name (terra)
  • The Claude model in mono (sonnet-4-6, opus-4-6, etc.)
  • A sage Live badge if the agent is deployed at Anthropic
  • A butter forked chip if it's a fork, a sage published · /slug chip if published
  • A description truncated to 2 lines
  • Footer: creation date, number of tasks launched, an Open/Install button, a delete button

Running an agent

Click Run. A panel opens with an optional input field — this is the context you want to pass to the agent for this specific run (for example a file to analyze, a URL, a specific question). If the agent is autonomous (cron, trigger), you can leave it empty.

The session starts, you see the agent's messages streaming live, and the final deliverable appears at the bottom. You can copy it, download it, or share it.

Sessions and history

Each agent run creates a session, with its own identifier, its inputs, its complete transcript, and its outputs. You can:

  • Resume a session in progress (the agent continues where it left off)
  • Replay a past session with the same inputs
  • Compare two sessions side by side (for debugging)
Session storageSession transcripts are stored at Anthropic, not with us. Forgent keeps only the metadata (date, status, message count). If Anthropic purges its sessions, we lose access to the details.

Deleting an agent

Click the kebab menu (the three dots) to the right of an agent, then Delete. The deletion is:

  • Immediate on the Forgent side — the blueprint and metadata disappear from your dashboard.
  • Asynchronous on the Anthropic side — we send the deletion order to Managed Agents, which finishes releasing the resources within a few minutes.
Deletion is permanentNo trash bin. No recovery. If you just want to temporarily disable it, instead delete the trigger in the Anthropic console and keep the agent on standby.

Importing local drafts

If you built agents in your browser before connecting a wallet, those drafts are saved locally. Once you sign in, a migration banner on My Agents offers to import them into your account in one click, so they follow you across devices. You keep full control: nothing is imported without your explicit approval, and existing entries are never overwritten.