omnara.com

# Omnara

> Omnara is an open-source alternative to Claude Managed Agents. It runs durable AI agents while you choose models, tools, machines, and the user interface. Use Omnara Cloud or self-host the Apache-2.0 agent platform.

## When to use Omnara

Build customer-facing or internal agents that need persistent state, tool execution, human approvals, or multiple machines. Omnara manages agent execution and state; your application controls its user experience. A single stateless model request may only need a model API.

Agents are API objects, not code deployments: launch one per customer, ticket, or thread with an API call, and change its behavior by posting a new config. Agents can use any compatible model and run on any machine, including sandboxes billed through Omnara, your own sandbox-provider account, or your own VMs, containers, and laptops through the Omnara daemon. Agent frameworks such as Vercel's eve, Mastra, or LangGraph suit a single agent built into an application's code; Omnara suits many agents managed through an API.

## Customers

Teams running agents on Omnara include [1Digital Agency](https://www.1digitalagency.com), [Stellon Labs](https://stellonlabs.com), [The Prompting Company](https://www.promptingcompany.com), [Modelence](https://modelence.com), [Meridian Supply](https://meridiansupply.org), and [Heavybit](https://www.heavybit.com).

> Omnara opened my eyes to what AI agents can do for our agency. We went from basic AI implementation with wizards and one-off prompts to more than ten real AI agents for proposal building, SEO analysis, sales, and R&D. We’re even developing agents for our customers. Now I have one place to run them all and see their costs.
>
> — Dan Kogan, Founder & CEO, 1Digital®

## Product and source

- [Pricing](https://www.omnara.com/pricing): Current hosted pricing and bring-your-own options.
- [Startups](https://www.omnara.com/startups): Credits for models and machines, and a direct line to the Omnara team.
- [Examples](https://docs.omnara.com/examples/overview): Agents you can deploy with one prompt to a coding agent.
- [Source code](https://github.com/omnara-ai/omnara): Open-source agent platform, Apache 2.0.
- [Quickstart](https://docs.omnara.com/quickstart): Launch your first agent.
- [Self-hosting](https://docs.omnara.com/self-hosting/deployment): Operate your own deployment.
- [API overview](https://docs.omnara.com/api/overview): Product API documentation.
- [OpenAPI specification](https://docs.omnara.com/api-reference/openapi.yaml): API operations, parameters, responses, and authentication requirements.
- [Authentication](https://docs.omnara.com/api/authentication): Personal and organization API keys and CLI device login.
- [OAuth server metadata](https://app.omnara.com/.well-known/oauth-authorization-server): Discovery for the Omnara CLI device authorization flow.
- [API catalog](https://www.omnara.com/.well-known/api-catalog): Links to the public API contract and documentation.
- [Agent resource discovery](https://www.omnara.com/.well-known/ard.json): Machine-readable directory of the API specification and setup skill.
- [Documentation index](https://docs.omnara.com/llms.txt): Detailed documentation in Markdown.
- [Agent setup skill](https://www.omnara.com/SKILL.md): Instructions for setting up Omnara and launching an agent.

Hosted billing and commercial integrations are separate from the open-source agent platform. Documentation and current pricing are authoritative for configuration and charges. The blog archive covers Omnara's previous remote coding product and is retained for historical reference. Its features and usage figures refer to that product. Omnara is now an open-source managed-agent platform.

## Guides

- [How Cursor's Agent SDK Works](https://www.omnara.com/guides/cursor-agent-sdk.md): How Cursor's Agent SDK and Cursor SDK work, what they cost, and what still runs on Cursor when you self-host.
- [How Vercel's eve Works](https://www.omnara.com/guides/vercel-eve.md): What Vercel's eve agent framework does well, what changes when you run it off Vercel, and what happens when you change an agent that's already running.
- [How to Build Durable AI Agents with Temporal](https://www.omnara.com/guides/temporal-ai-agents.md): How to make AI agents durable with Temporal: the OpenAI Agents SDK and Pydantic AI integrations, running Claude Code on it, writing your own loop, and costs.
- [How Vercel's durable-harness Works](https://www.omnara.com/guides/vercel-durable-harness.md): How Vercel's durable-harness keeps Claude Code, Codex, OpenCode, and Pi running through crashes, which of their features you lose, and when to use something else.
- [How to Run Vercel AI SDK Agents in Production](https://www.omnara.com/guides/vercel-ai-sdk-agents-production.md): Run Vercel AI SDK agents in production: ToolLoopAgent vs WorkflowAgent, approvals, resumable streams, Workflow pricing, self-hosting, and a crash test.
- [How to Build Subagent Systems](https://www.omnara.com/guides/build-subagent-systems.md): What subagents are, when they help or hurt, how mixing models from different labs changes quality and cost, and a worked example: a website auditor with judges from four labs. Covers Claude Code, Codex, Cursor, and Omnara.
- [How to Deploy Durable Agents with Pi Durable](https://www.omnara.com/guides/deploy-pi-durable-agents.md): Deploy Pi Durable agents on Cloudflare Durable Objects, Vercel's durable-harness, or your own servers. What Cloudflare costs, what you still build yourself, and how it compares with Omnara.
- [How to Build Your Own AI Code Review Bot](https://www.omnara.com/guides/build-ai-code-review-bot.md): Build your own AI pull request reviewer with any agent harness: cheap reviewer subagents from different providers, a stronger model that checks their findings, and one GitHub review. Prompts and setup included.
- [Cloud Agent APIs Compared in 2026](https://www.omnara.com/guides/cloud-agent-apis.md): Compare the APIs for launching cloud and background agents from your own code: Cursor, GitHub Copilot, OpenAI, Claude, Google, Devin, and Omnara, including which ones you can use for your own users.
- [Self-Hosted Managed Agents in 2026](https://www.omnara.com/guides/self-hosted-managed-agents.md): What each self-hosted managed agent option runs on your infrastructure and what stays with the vendor: Claude Managed Agents, Claude Code runners, OpenAI's Agents API, AWS, Cursor, Devin, and open-source platforms.
- [Claude Managed Agents Alternatives in 2026](https://www.omnara.com/guides/claude-managed-agents-alternatives.md): Compare 7 Claude Managed Agents alternatives for 2026: OpenAI's Agents API, Gemini, LangChain, DigitalOcean, AWS, Azure, and open-source Omnara.
- [How to Deploy an AI Agent to Production in 2026](https://www.omnara.com/guides/deploy-ai-agents-to-production.md): What changes when an AI agent serves real users: sessions, sandboxes, per-user credentials, approvals, cost limits, and tracing, plus what happened when we crashed an agent mid-task.
- [Four Ways to Build an AI Agent in 2026](https://www.omnara.com/guides/agent-framework-vs-harness-vs-runtime.md): What workflow builders, agent frameworks, harnesses, and runtimes are, when to use each, and the main options in each, from n8n and LangGraph to the Claude Agent SDK.

## Blog articles

- [How we ran Terminal-Bench 4.0 11% cheaper and 12% faster than the frontier labs](https://www.omnara.com/blog/terminal-bench-cheaper-faster.md): Benchmarking Omnara with Terminal-Bench 4.0
- [Agent Memory Is in the Wrong Place](https://www.omnara.com/blog/agent-memory-is-in-the-wrong-place.md): Why agent memory belongs in the control plane, not a sandbox, and how we still gave agents real files, ripgrep and sed.
- [The Harness Doesn't Matter](https://www.omnara.com/blog/the-harness-doesnt-matter.md): What an AI agent harness really is, why most harnesses are equivalent, and where they can still meaningfully differ.
- [Serverless Agents](https://www.omnara.com/blog/serverless-agents.md): No single machine should own the agent loop. Put agent continuity in a persistent control plane, make the loop serverless, and expose machines as tools.
- [The Log Is the Agent](https://www.omnara.com/blog/the-log-is-the-agent.md): An agent is its durable event history: the log that lets it resume, fork, migrate, and survive failures across models, runtimes, and machines.
- [What Is an Async Agent, Really?](https://www.omnara.com/blog/what-is-an-async-agent-really.md): Spoiler: you may or may not find out

## Optional

- [Blog](https://www.omnara.com/blog)
- [Guides](https://www.omnara.com/guides)
- [RSS feed](https://www.omnara.com/feed.xml)
- [Sitemap](https://www.omnara.com/sitemap.xml)
- [Privacy](https://www.omnara.com/privacy)
- [Terms of Service](https://www.omnara.com/terms)
- [Trust Center](https://trust.omnara.com): Security and compliance documentation.
- [Contact](mailto:contact@omnara.com)