Agent Manager by HF

Your coding agents, together in one private workspace.

Run Claude Code, Codex, Gemini CLI, opencode, Hermes, OpenClaw and fx from your browser. Follow several tasks side by side, read the conversation or open the terminal, and pick up from another device while your Space stays running.

Get started · Development · Report an issue

The current interface with example conversations.

• Keep several tasks in view. Group sessions by project, arrange them side by side, and use Overview to see who is working and whose reply needs your attention.
• Read at the level you need. Switch between the live terminal and a conversation reader with formatted replies, expandable tool activity, search, and earlier history. Reader support depends on the agent's transcript format.
• Work with your files. Browse workspace folders, preview and edit files, upload documents or screenshots, and keep a shell beside your agents.
• Give agents shared context. Manage reusable skills in one place and let agents discover, message, and wait for other sessions through the local API.
• Keep the agents current. Update each CLI fromSettings → General → Agents without rebuilding the Space.
• Check usage and schedule work. See token usage, estimated costs, and available quota information; schedule prompts and enable device notifications. Available usage details vary by provider.
• Bring another machine. Connect an agent running on your laptop or a remote server to the same workspace's conversation view.

Each session has its own conversation. Sessions can share a folder when you want them to work on the same project; groups organize the interface and do not isolate files or credentials. Use your own provider login or API keys.

You need a Hugging Face account and access to at least one supported coding agent. The app runs in a Docker Space; a private storage bucket mounted at /data keeps your files, logins, and conversation history across restarts.

Open the Agent Manager template and choose**⋮ → Duplicate this Space** . Set visibility toPrivate . 2. In your new Space's Settings → Storage Buckets , create or select aprivate bucket, mount it at**/data** , and useread-write access. Do this before signing in to agents. See theHugging Face storage guide for the mount controls. 3. Once the Space is running, open the app. The first-run window lists the agents in your image, shows which are ready, and walks you through connecting one — either by opening a session and following its sign-in, or by adding an API key as a Space secret. Reopen it fromSettings → General → Show guide .For Codex , chooseSign in with Device Code in the session. Its ChatGPT option redirects tolocalhost on the machine the agent runs on, which your browser cannot reach from a Space. 4. Send your first task. Add another session to review it, or create a group to keep both in view. Overview brings you back to all your sessions.

Keep both the Space and its bucket private. Agent Manager has no login of its own: anyone with access to the app can use its shell and logged-in agents. The public template is an installation page. The privacy lock helps detect an exposed instance.

Agent Manager is open source under Apache 2.0. Your agent subscriptions or API usage, Space hardware, and storage follow their providers' billing.

Set up the Space and bucket with Python instead #

Install the Hub client and sign in with an account that can create Spaces and buckets in your namespace:

pip install -U huggingface_hub
hf auth login

The names come from the account you just signed in as:

from huggingface_hub import HfApi, Volume

api = HfApi()
username = api.whoami()["name"]
space_id = f"{username}/agent-manager"
bucket_id = f"{username}/agent-manager-data"

api.create_bucket(bucket_id, private=True, exist_ok=True)
api.duplicate_repo(
    from_id="lvwerra/agent-manager-template",
    to_id=space_id,
    repo_type="space",
    private=True,
    space_volumes=[
        Volume(type="bucket", source=bucket_id, mount_path="/data"),
    ],
)

Open https://huggingface.co/spaces/<your-username>/agent-manager when the build finishes, then continue from step 3 above. To attach a bucket to an existing Space, use the same Settings controls or the Hub volume API.

Closing the browser or switching devices does not stop your agents. They run in the Space's backend as long as the Space stays awake.

A Space sleep, restart, or rebuild ends running processes. With the bucket mounted, saved workspace files and checkpointed CLI state survive. Agent Manager can restart eligible sessions and resume conversations where the CLI supports it; it cannot preserve a running shell command through a reboot. Configure this in Settings → General → Restart sessions after a reboot.

Without a bucket, local files and logins are temporary. For details on what is saved, see state checkpoints and bucket backups.

Shared Codex sessions are a work in progress: a draft proposal describes the intended Agent Manager, TUI and Codex Remote workflow, with an experimental read-only preview and an isolated binding pilot. Shared creation and automatic migration are not enabled by default, and none of this changes how you sign in to Codex today.

Contributions and bug reports are welcome on GitHub. Include your agent type, browser, and steps to reproduce when reporting a problem. For a security issue, see SECURITY.md.