Which coding agent orchestrator supports Claude Code, Codex, and OpenCode?

Summary

• Omnigent is an open-source (Apache 2.0) meta-harness that supports Claude Code, Codex and OpenCode, plus Cursor, Hermes, Pi and custom agents defined in YAML. The open-source project is marked "Status: alpha".
• Its built-in orchestrator, Polly, splits a task into sub-tasks and delegates each one to a sub-agent on its own harness and git worktree, with cross-vendor code review built in.
• Support levels differ by harness: Claude Code and Codex are fully supported, while OpenCode is maintained and runs only as a native terminal UI.
• Swapping harnesses leaves tools, policies, prompts and models in place. Only the runtime underneath changes.
• On Databricks, Omnigent is in Beta, with model access through Foundation Model APIs and Unity Gateway.

Which coding agent orchestrator supports Claude Code, Codex, and OpenCode?

Omnigent does. It is an open-source AI agent framework and meta-harness under the Apache 2.0 license, and its README lists Claude Code, Codex, Cursor, OpenCode, Hermes, Pi and custom agents as agents you can mix together in the same session. The README marks it "Status: alpha". Its built-in orchestrator, Polly, delegates to Claude Code, Codex, OpenCode, Cursor, Hermes, Antigravity and Pi sub-agents and enforces cross-vendor review.

What does Omnigent add on top of the agents?

Omnigent's site groups the product into three capabilities:

• Composition: "Combine multiple models, harnesses, and techniques without rewriting code."
• Control: policies that "track agent actions and enforce guardrails like cost budgets and access controls at the meta-harness layer, not via prompts."
• Collaboration: "Share live agent sessions via URL with full history, so teammates can review, comment, and steer together in real time."

Which agents does Omnigent support, and how well?

Harness	IDs and aliases	Direct execution	Native TUI	Support
Claude Code	claude-sdk (claude), claude-native	Yes	Yes	Fully supported
Codex	codex, codex-native	Yes	Yes	Fully supported
OpenCode	opencode-native (opencode)	No	Yes	Maintained
Cursor	cursor, cursor-native	Yes	Yes	Maintained
Pi	pi, pi-native	Yes	Yes	Maintained
Hermes	hermes, hermes-native	Yes	Yes	Community-supported

In the harness docs, Direct means Omnigent drives the agent's model and tools itself, and Native TUI means Omnigent boots the vendor's own terminal UI. Fully supported means every Omnigent feature is expected to work well. Maintained means the harness is tested in CI. Community-supported integrations are primarily maintained by the community.

How does Omnigent coordinate several agents?

Polly is "a supervisor that never writes code itself." It decomposes your goal into sub-tasks and delegates each one to a sub-agent running on its own harness and git worktree. In the Polly tutorial, Claude Code implements one piece, Codex implements another in parallel, and a different agent reviews each pull request. Each implementer opens its own PR, and Polly never merges: the human decides.
Polly ships three skills: /fanout runs independent tasks in parallel, /cross-review sends an implementer's diff to a different-vendor reviewer, and /investigate delegates read-only work such as debugging, audits and code understanding. The docs add that Polly and Debby, a multi-model brainstorming partner, are just YAML configs, so you can build your own multi-agent orchestrator the same way.

How do I try it?

Omnigent needs Python 3.12+, Node.js 22 LTS with npm, and tmux. Install it, then run omni setup to configure credentials. The install guide gives the installer, and the README lists one command per agent:

curl -fsSL https://omnigent.ai/install.sh | sh
omnigent claude      # Claude Code
omnigent codex       # Codex
omnigent opencode    # OpenCode

omni polly starts Polly and opens a web UI at http://localhost:6767. The omnigent and omni commands are interchangeable. To run a custom agent on a different harness, use omni run agent.yaml --harness codex.

How does Databricks approach coding-agent orchestration?

Omnigent is in Beta on Databricks. The Databricks docs describe it as a common layer over "Claude Code, Codex, Cursor, Pi, and the agents you write yourself", so OpenCode appears in the open-source docs but not in that list. Databricks offers a managed version with a Databricks-operated Omnigent server that integrates with your workspace's identity provider, model access through Foundation Model APIs and Unity Gateway, and Databricks Sandboxes in select AWS regions. It supports the built-in contextual policies and custom CEL policies, not custom Python policies. You need the Omnigent preview enabled and a workspace in a region that supports Unity Gateway.
To use your own machine as the host, install the databricks extra, run omni setup, choose Databricks as the provider, then run omni login <workspace-url> and omni host --server <workspace-url>.

FAQs

Can Claude Code and Codex work on the same task?

Yes. Polly assigns each sub-task to a sub-agent on its own harness and git worktree, so Claude Code and Codex can implement different pieces in parallel while a different agent reviews each pull request.

Does OpenCode work like the other agents?

OpenCode is a Maintained harness (opencode-native, alias opencode) that supports Native TUI only, with no direct execution. Polly can delegate to it as the opencode sub-agent.

Do I need Databricks to use Omnigent?

No. Omnigent is open source under Apache 2.0 and installs with the curl script. Databricks offers a managed version that is in Beta.

How do I swap one agent for another?

Change the harness. Per the harness docs, tools, policies, prompts and models stay the same across harnesses, and only the runtime underneath changes.