Can I use my own models and runtime with OpenHands?

Yes. OpenHands is built around the principle of "Your agents. Your models. Your runtime." — the platform explicitly supports bringing your own model and running in your own environment. You can start locally on your own machine, or use OpenHands Cloud if you'd rather not manage infrastructure. The trade-off is between control (local/self-managed runtime, your own model choice) and convenience (Cloud, managed hosting).

What "your models" means in practice

OpenHands describes itself as model-agnostic: the platform lets you use any model rather than locking you into a single provider. In the interface, model selection appears as a configurable option ("Model" alongside "Open Workspace" and "Plugins"), which signals that the model is a setting you control rather than a fixed part of the product.

This matters if you:

• Already have access to a specific model (commercial API or self-hosted) and want to keep using it
• Need to switch models depending on the task (for example, a faster model for triage, a stronger one for code review)
• Have cost, latency, or data-residency reasons to avoid a default provider

The page does not list specific supported model names or providers, so treat "any model" as the platform's stated positioning and verify your particular model against current documentation before committing.

Runtime options: local vs. Cloud

OpenHands gives you two starting points, and the choice affects how much you manage yourself.

Both paths share the same core idea — your agents, your models, your runtime — but local means you own the runtime setup, while Cloud means OpenHands handles it for you. The page presents local as the entry point ("Start locally") and Cloud as the scale-up path, with enterprise controls mentioned for larger deployments.

Integrations: where agents get triggered

The runtime and model are only half the picture. OpenHands is designed to plug into the tools your team already uses, so agents start work from real signals rather than manual prompts. The page names these integration points:

• GitHub — trigger workflows on pushes and pull requests
• Slack — watch channels for @openhands mentions and reply when the agent finishes
• Jira — triage incoming incidents and bugs
• Linear — classify new issues, suggest labels, find duplicates
• Schedules — run automations on a fixed cadence (for example, daily or weekday mornings)

This is what makes the "bring your own" model practical: the agent runs in your environment with your model, but it's still wired into the systems where work actually arrives.

What you can automate once it's connected

The page ships with recommended automation templates that show the intended use cases. A few concrete examples:

• Slack channel monitor — watch a channel for mentions and open a conversation with the message context
• Slack standup digest — collect standup replies and post a weekday digest to the team channel
• PR review — review newly opened pull requests and post findings as inline comments
• PR triage digest — review new PRs in a repo, flag risky changes, and post a priority-ordered summary (example shown: weekdays at 09:00 on acme/frontend-app)
• Nightly security pass — scan a repo for known vulnerabilities, outdated dependencies, and insecure patterns, then produce a prioritized remediation summary (example: daily at 01:30 on acme/backend-api)
• Docs sync on push — run on every push

These templates are starting points; the page frames the platform as "build it once, let agents keep it running," with the repetitive parts of the SDLC on autopilot.

How to decide

Choose local if you need your code and model calls to stay on your own infrastructure, or if you're evaluating the platform before rolling it out. Choose OpenHands Cloud if you want the automations running without managing a runtime yourself. In both cases, confirm that your specific model is supported and that the integrations you need (GitHub, Slack, Jira, Linear) match your stack — the page lists these as supported, but your exact workflow may need verification.

One thing the page does not settle: pricing and any login requirements for Cloud. A pricing link exists, but no figures are given here, so check that page directly before assuming cost or access terms.