Can I share a Claude Code session with a teammate?

Summary

• Yes, through Omnigent. Omnigent is an open-source meta-harness that runs Claude Code as a harness inside a shareable session. Start it with omnigent claude, click Share in the web UI, and send the link. Teammates watch the session in real time and, if you grant it, drive the agent.
• Sharing has two permission levels. Read lets a teammate follow along. Edit lets a teammate send messages and drive the agent, which acts on the machine that hosts the session, so grant it carefully.
• You can bring in a chat you already started. omnigent import --harness claude --session <session-id> converts a local Claude Code session into a regular Omnigent session that gains the web UI, persistence, and collaboration.
• Check the stage. The open-source project's README lists its status as alpha. The Databricks managed version is in Beta, and a workspace admin must enable the preview.

Can I share a Claude Code session with a teammate?

Yes, if you run Claude Code through Omnigent. Omnigent is multi-user: once your server is reachable by your team, you can share a live session with a link, let a teammate co-drive, or fork the conversation so someone can continue independently. Teammates see messages, tool calls, and file changes stream in real time, and the same session works across the terminal, web UI, and mobile at once.

How do I start a Claude Code session that a teammate can join?

1. Run omnigent claude. The README labels it "Claude Code, in a session your team can join". The shorter omni command is interchangeable with omnigent.
2. Make sure teammates can reach your server. A local server is reachable only on your network, so for anyone off it, deploy an always-on host. On Databricks, you open sessions at <workspace-url>/omnigent.
3. In the web UI, click Share on the session and send the link. Teammates open it in their browser and join in real time.

What can a teammate do in a shared session?

Option	What it allows
Read	Follow the session stream. In the open-source docs, a Read share shows messages, tool calls, and streaming updates but hides workspace files unless the owner turns on the Workspace files toggle.
Edit	Send messages and drive the agent. The Databricks docs describe Edit as full control of the session: send messages, open shells, edit files in the host file system, and rename the session.
Co-drive from a terminal	Attach to a running session with omnigent attach <session_id>. The teammate's messages run on the machine that hosts the session.
Fork	Clone the conversation and continue independently from the fork point with omnigent run --fork <session_id>. The original session is unaffected.

How does sharing work on Databricks?

On Databricks, Omnigent (Beta) uses your workspace identity, so there is no separate Omnigent login, account, or password. Sessions appear in the <workspace-url>/omnigent UI and can be shared with other workspace users, whether they run on a Databricks Sandbox or on your own registered machine. Each session is private to its owner until shared, and workspace admins do not have blanket access to every session. Owners, collaborators, and the author of each turn are shown by their workspace email. A workspace admin must enable the Omnigent preview, and the workspace must be in a region that supports Unity Gateway.

What are the risks of giving a teammate Edit access?

Co-driving grants code execution on the host. The open-source docs say to share Edit only with people you trust to run arbitrary code on your machine, because a teammate with Edit gets the same access to your filesystem, credentials, and network that the session has. The Databricks docs warn that an editor can run shell commands and modify files on the host through the agent, and can also view your other session chats.
To collaborate more safely, share read-only, host the session on a cloud sandbox host so the agent runs in an isolated remote container, or apply an OS sandbox policy. The open-source docs add that everything an Edit collaborator does goes through the agent, so it passes the same policy and approval gates as the owner's own turns.

How do I share a Claude Code chat I already started?

Import it first. omnigent import --harness claude --session <session-id> imports one chat by its harness-native session ID, and --last N imports your most recent chats. The docs say the source transcript is converted to ordinary Omnigent items and stored as a regular session, so it gains the web UI, persistence, and collaboration. A given source session can be imported once unless you pass --force. Then share the imported session like any other.

FAQs

Can a teammate see my local files?

With a Read share, not by default. The open-source docs say view-only viewers do not see the Files, Changes, or GitHub diff surfaces unless the owner opens the Share dialog and turns on Workspace files. Edit collaborators always have workspace access.

Can a teammate answer the agent's permission prompts?

Only with Edit. The open-source docs say pending prompts, such as approvals and questions, are answerable by anyone who can drive the agent, and a Read share can neither see nor answer them.

How do I tell who said what in a shared session?

In the open-source version, once two or more people have contributed, Omnigent prefixes each message the model sees with the author's account in square brackets, so the agent can tell collaborators apart. On Databricks, owners, collaborators, and the author of each turn are shown by workspace email.

Is this generally available?

No. The open-source project's README lists status alpha, and the Databricks managed version is in Beta. See the Omnigent collaboration docs and the Databricks Omnigent docs.