Omnigent logopeterindia.net
Knowledge Hub

Omnigent

Omnigent is an open-source meta-harness: a common layer over Claude Code, Codex, Cursor, OpenCode, Hermes, Pi and custom agents that adds policies, sandboxes and shared live sessions. This guide covers what it does, its alpha status and licence, how to install it, and the sources behind each claim.

28 topics•4 sections•Updated November 2026

What Omnigent Is

6 entries

Primary sources for the project: its site, repository and installation guide. Descriptions follow the project's own pages.

01Omnigent siteOmnigent describes itself as an open-source meta-harness: a common layer for composing, governing and collaborating on AI agents on top of the harnesses you already use. 02Omnigent repository (GitHub)Source code under the Apache-2.0 licence. The README calls it the open-source meta-harness for all your AI agents and marks the project alpha; GitHub lists v0.20.0 (28 October 2026) as the latest release. 03Install OmnigentInstallation with the install script, uv, pip or Homebrew. The README says Python 3.12 or newer is required. 04Omnigent on PyPIThe omnigent package, described as a declarative agent authoring and runtime framework; PyPI lists 0.20.0 (28 October 2026). 05ReleasesRelease notes for each version. 06API referenceREST API and SDK reference.

Features

6 entries

The capabilities the project documents: running several harnesses, policies, sandboxes and collaboration.

07Coding agents guideLaunching harnesses such as Claude Code and Codex through Omnigent. The README lists Claude Code, Codex, Cursor, OpenCode, Hermes, Pi and custom agents defined in YAML. 08Built-in agents (Polly and Debby)Polly, a coding orchestrator, and Debby, a model debate agent, plus how to build your own agents in YAML. 09Contextual policiesPolicies intercept actions such as tool calls, LLM requests and file operations and decide to allow, ask for approval or deny. They are stateful, so a cost budget can track spend across a session. 10Secure sandboxes (Omnibox)Local OS sandboxes and remote cloud sandboxes with filesystem and network limits and credential brokering. The README names providers such as Modal, Daytona, E2B and Kubernetes. 11Homebrew tapThe Homebrew formula repository for installing Omnigent on macOS and Linux. 12Omnigent community (Discord)The project's community chat, linked from its README.

Articles and Coverage

6 entries

The project's own blog posts and trade-press coverage of the launch. Blog posts are vendor-authored.

13Meta-Harness: the Missing Layer Above Claude Code, Codex, and Pi (Omnigent blog)The Omnigent team's explanation of the meta-harness idea. 14Allow, ask, deny: contextual policies in Omnigent (Omnigent blog)Introduction to contextual ALLOW/ASK/DENY guardrails, with built-in, custom Python and CEL examples. 15Stop Surprise LLM Bills With Production-Ready Cost Controls (Omnigent blog)How Omnigent keeps spend predictable: routing, progressive budgets, custom pricing for self-hosted models and per-model usage tracking. 16Moving agent orchestration from code to configuration (Omnigent blog)Defining an orchestration setup in YAML and letting Omnigent handle sessions, messaging, isolation and policy enforcement. 17Databricks launches open-source Omnigent for AI agents (IT Brief)Trade-press report on the open-source release as a single control layer for fragmented AI agent tools, with governance and cost controls. 18Databricks open-sources Omnigent (MarkTechPost)News report on the release and the harness versus meta-harness distinction; the site blocks automated checks.

Related Pages

10 entries

Related guides on this site.

19A Meta-Harness for AI AgentsThe meta-harness idea, Omnigent as the reference example, related tools and research. 20Agent HarnessesThe five memory-generation harnesses on the FrontierMemory v1.1 leaderboard. 21AI Coding AgentsDirectory of coding agents such as Claude Code and Codex. 22AI Agent Orchestration: Harness-Centric ApproachAgent harness definitions, research and SDKs. 23AI Agent GuardrailsFive guardrail checkpoints for an agent. 24AI Model RoutersThe routing layer that directs requests across models. 25AI Agent Management PlatformsThe 2026 Gartner Magic Quadrant for the category and the vendors named in it. 26Agentic AIThe Agentic AI hub. 27AI Agent Governance PlatformsDiscovering, observing and governing autonomous AI agents. 28AI & LLM GatewaysGateways that route, govern and secure LLM, MCP and agent traffic.

Frequently Asked Questions

What is Omnigent?

Omnigent is an open-source meta-harness: a common orchestration layer over agent harnesses such as Claude Code, Codex, Cursor, OpenCode, Hermes, Pi and custom agents defined in YAML. It lets you swap or combine harnesses without rewriting, enforce policies and sandboxing, and share live sessions from a terminal, browser, phone or desktop app.

Who makes Omnigent?

Trade press, including IT Brief and MarkTechPost, reports that Databricks released Omnigent as open source in June 2026. The source code lives in the omnigent-ai organization on GitHub, which created the repository on 11 June 2026.

What licence and status does Omnigent have?

The repository is licensed under Apache-2.0, and its README marks the project alpha, so interfaces and behaviour may change. Evaluate it in a test environment before relying on it.

Which harnesses does Omnigent support?

The README lists Claude Code, Codex, Cursor, OpenCode, Hermes, Pi and custom agents defined in YAML, and mentions native wrappers and SDK harnesses for others. Check the project's harness documentation for the current list and versions.

What are Omnigent policies?

Policies intercept every action, such as tool calls, LLM requests and file operations, and decide in real time to allow, ask for approval or deny. They are stateful, so they can enforce cumulative cost budgets or escalation based on what has happened in the session, and they can apply to the whole server, one agent or a single chat.

What are Omnigent sandboxes?

Omnigent can run agents in local OS sandboxes or in remote cloud sandboxes, restricting filesystem and network access and brokering credentials on Linux and macOS. The README names providers including Modal, Daytona, E2B, Kubernetes and Databricks.

Does Omnigent run on Windows?

The README says Omnigent runs natively on Windows in a degraded mode: the server, web UI and SDK-based harnesses work, but the native terminal wrappers and the bwrap or seatbelt filesystem and network sandboxing are not available. It recommends Linux, macOS or WSL for those.

How do I install Omnigent?

The README gives a one-line install script, or uv, pip or Homebrew, and requires Python 3.12 or newer; some features need Node.js and tmux. See the project's install page for current commands.

How is Omnigent different from an agent framework or an AI gateway?

Agent frameworks help you build agents, and AI gateways route and govern model and tool traffic. Omnigent mainly runs and controls agents and harnesses that already exist, adding policies, sandboxes and shared sessions, though it can also define custom agents in YAML. See the AI Agent Frameworks and AI Agent Gateways pages.

How current is this page?

Reviewed on 4 November 2026. Facts come from Omnigent's repository, PyPI listing and documentation (v0.20.0, released 28 October 2026) and from trade-press reports of the June 2026 launch. All outbound links returned 200 except the MarkTechPost article, which blocks automated checks, and the IT Brief article, which loaded earlier on 3 November but timed out on the final check. The project is alpha and changes quickly, so confirm details on its own site.