Can I move a Claude Code conversation to Codex?

Summary

• Yes. With Omnigent, an open-source meta-harness, you can bring a Claude Code chat into Omnigent and clone it to Codex. The Omnigent tutorial says the new agent "picks up the full conversation history and keeps going".
• The steps are omnigent import --harness claude --session <session-id>, then Clone session in the web UI and pick Codex. A session you started with omni claude is already an Omnigent session, so you only clone it.
• A session's bound agent can't be changed in place, so moving to Codex always creates a copy, and the original session stays as it was.
• Claude Code and Codex are both in Omnigent's "Fully supported" harness tier.
• The open-source repository says "Status: alpha", and on Databricks "Omnigent is in Beta".

Can I move a Claude Code conversation to Codex?

Yes, by bringing the conversation into Omnigent and cloning it. Omnigent is "an open-source meta-harness" that gives you a common orchestration layer over Claude Code, Codex, Cursor, OpenCode, Hermes, Pi, and the agents you write yourself. Its tutorial says: "Started in Claude Code? Clone the session and continue the copy with Codex (or vice versa)." The copy continues from the full conversation history, and the original session is not affected.

How do you move the conversation, step by step?

1. Get the chat into Omnigent. A chat you started with omni claude is already an Omnigent session. For an existing local Claude Code chat, run omnigent import --harness claude --session <session-id> with the harness-native session ID. Use --last N instead to import your N most recently modified chats, up to a maximum of 100.
2. Or import from the web UI. Open Settings > Import sessions, pick an online machine, and choose recent sessions or a session by ID. If no machine is online, start one with omnigent host.
3. Clone the session. In the web UI, choose Clone session and pick Codex as the agent for the copy.
4. Continue in Codex. "The new agent picks up the full conversation history and keeps going."

The docs say an imported transcript "is converted to ordinary Omnigent items and stored as a regular session". A given source session can be imported only once unless you pass --force.

Why is it a clone and not an in-place switch?

The tutorial explains: "Sessions in Omnigent belong to you, not to a specific agent, so a copy doesn't have to continue with the agent it started with." Forking from the CLI with omni run --fork <conversation_id> keeps the same agent, so use Clone session in the web UI to pick a new one.
The programmatic docs add that "a session's bound agent can't be switched in place" and that POST /v1/sessions/{id}/switch-agent returns 410 Gone. In the API, a fork request accepts an agent_id to choose the new agent. The v0.17.0 release notes (2026-10-06) say clients that switched a session's agent in place "must fork the session into another agent instead".

What does the Omnigent documentation promise?

Question	What the docs say
Does Codex see the earlier conversation?	The new agent "picks up the full conversation history and keeps going"
Is the original session changed?	A fork works "without affecting the original"
Can I switch the agent in place?	No. A session's bound agent can't be changed in place
Can I import the same chat twice?	Not by default. Pass --force to replace the previous import
Is anything beyond the history documented as carried over?	The tutorial and docs describe the conversation history being copied and list no other state, so check your project files and tool setup in the new session

What does the one-line YAML change do?

For an agent you define in YAML, you select the harness with executor.harness, for example harness: codex, or override it at run time with omni run agent.yaml --harness codex. The docs say "Tools, policies, and other config stay the same across harnesses." To continue an existing conversation on another agent, the documented route is Clone session.

What does Databricks add?

Omnigent is open source under Apache 2.0. Databricks offers a managed version with a Databricks-operated Omnigent server integrated with your workspace identity provider, model access through Foundation Model APIs and Unity Gateway, and Databricks Sandboxes for collaborative agent coding in select AWS regions. The Databricks docs list requirements and limits: the Omnigent preview must be enabled for your workspace, your region must support Unity Gateway, custom policies use Common Expression Language (CEL) and not custom Python policies, and Databricks Sandbox hosts route model access through Unity Gateway without bring-your-own API keys.

FAQs

Can I use Omnigent with Claude Code and Codex right now?

Yes. Run omni claude or omni codex. The docs say each command "launches a built-in agent" with no registration step. Both harnesses are listed as "Fully supported", defined as "Every Omnigent feature is expected to work well with these harnesses."

Can I import a Codex chat the same way?

Yes. omnigent import --harness codex --session <session-id> imports a Codex chat, and omnigent import --harness all --last 25 imports the 25 most recent chats across all supported harnesses.

Can multiple people use the same Omnigent session?

Yes. The README describes co-driving: a teammate co-attaches to your running session with omnigent attach <session_id>, and their messages execute on your machine.

How mature is Omnigent?

The open-source repository carries a "Status: alpha" badge. On Databricks, "Omnigent is in Beta", and the preview must be enabled for your workspace.