EveStudioDocs

The workspace

Once an agent is active, Studio's window has three regions: the left rail (your agents and their chat threads), the per-agent header (name, status, and run/deploy controls), and the top tab bar (everything you do with the agent). This page is the map; each tab links to its own doc.

The left rail

The rail lists every agent Studio knows about under an Agents heading. At the top:

  • Add existing: opens a folder picker to register an Eve app you already have (the folder containing agent/).
  • + ("Create a new agent"): scaffolds a new agent. See Create an agent.

Click an agent to make it active. It expands to show that agent's chat threads, with an inline New chat to start a fresh one. Threads are your saved conversations; see Chat.

Managing threads:

  • Hover a thread to reveal an Archive icon; archiving moves the thread out of the sidebar without deleting it.
  • An Archived · N row appears when an agent has archived threads. Click it to open the Archived sessions modal, where you can Reopen a thread back into the rail or delete it permanently.

Settings sits at the bottom of the rail, under the updater badge when an update is available. Its About section states that Eve Studio is an independent open-source project by Kybernesis, not an official Vercel product, and that agents need eve 0.49+ and Node 24.

The per-agent header

Above the tabs, the header shows the active agent's state and its controls:

  • Name of the agent.
  • A status pill: stopped, starting, running, or error.
  • A mono metadata line in three segments separated by |:
    • eve 0.49.0: the eve version installed in the agent. When npm has a newer release, an ↑ 0.49.1 available chip follows it; click it to upgrade in place (see Keep eve current). When eve info reports discovery errors, a red N build errors chip appears; click it to open Capabilities.
    • Vercel · linked · deployed 14d ago my-agent.vercel.app, Vercel · linked · not deployed, or Vercel · not linked.
    • local · :2000 while the dev server runs, or local · starting…, local · stopped, local · error.
  • Start / Stop: starts or stops the local dev server (eve dev). Local Chat needs this running. If a dev server for the agent is already up, Studio adopts it instead of starting a second one. While running, a small link icon opens the dev server's info route in your browser.
  • Deploy: ships the agent to production (eve deploy). On an agent that isn't linked yet the button reads Link & deploy and opens the Deploy tab's Environment sub-tab, where the link flow lives. See Deploy to Vercel.

The top tab bar

The tabs are the surfaces for building, running, and shipping the agent:

  • Chat: talk to the agent's local dev server or its production deployment, with streamed tool calls, subagent delegations, reasoning, approvals, and session controls (Cancel, Compact, Clear, Reset).
  • Instructions, with Prompt and Model sub-tabs: edit instructions.md (the system prompt) and the model and reasoning in agent.ts.
  • Capabilities, with Tools, Skills, Subagents, and Hooks sub-tabs: browse, create, open, edit, and delete each; enable self-modification.
  • Integrations, with Connections and Channels sub-tabs: install from the eve registry, run guided channel wizards, and manage connections and Vercel Connect connectors.
  • Memory: Eve's session memory, the agent's memory slots, and Kybernesis Arcana for long-term memory, with a brain browser.
  • Schedules: list, create, and test cron jobs that wake the agent (agent/schedules/*.ts).
  • Deploy, with Deploy & Logs, Environment, and Sandbox sub-tabs: build, deploy, view logs, manage environment variables, and inspect the sandbox.
  • Evals: inspect and run the agent's evaluations.

Studio reads what the tabs show from Eve's compiled manifest and refreshes it with eve info --json (which regenerates Eve's discovery and compiled manifest without booting a server) whenever the manifest is older than the files under agent/, or when you hit a tab's reload button.

Next