AI Coding Agent Platforms
How do OpenHands Cloud pay-as-you-go credits work if I use the OpenHands at-cost LLM provider?
6 min read
Quick Answer: In OpenHands Cloud, pay-as-you-go credits are simply a prepaid balance you consume when you use the OpenHands at-cost LLM provider. You’re billed based on actual token usage at pass-through or near pass-through rates, and your credits only apply when you use OpenHands’ at-cost LLMs—not when you bring your own LLM key.
Why This Matters
If you’re running agents against real repositories, CI, and ticket queues, you don’t want vague “AI minutes” or opaque bundles. You want to know exactly how your OpenHands Cloud pay-as-you-go credits behave when you choose the OpenHands at-cost LLM provider: what triggers spend, how it scales with token usage, and how it fits alongside BYOK (bring-your-own-key) models. Clear mechanics here lets you forecast costs as you scale from one developer to an org-wide rollout.
Key Benefits:
- Predictable token-based billing: Credits map directly to token usage with the OpenHands at-cost LLM provider, so you can model spend from your workflows and codebase size.
- Flexible deployment choices: Use pay-as-you-go credits for OpenHands’ at-cost LLMs while still keeping the option to bring your own LLM keys for other workloads.
- Scales with real engineering work: As you spin up more agents for PR summaries, test fixes, or dependency upgrades, your credit consumption grows linearly with the tokens those agents actually use.
Core Concepts & Key Points
| Concept | Definition | Why it's important |
|---|---|---|
| OpenHands Cloud pay-as-you-go credits | A prepaid or usage-based balance in OpenHands Cloud that is consumed when you call the OpenHands at-cost LLM provider. | Gives you a clean, token-based cost model for hosted LLM usage without committing to a fixed seat/license bundle. |
| OpenHands at-cost LLM provider | A hosted LLM option managed by OpenHands, priced at or near pass-through rates per token. | Lets teams start fast without managing their own LLM account, while still keeping costs tied to real token consumption. |
| BYOK (bring your own LLM key) | Using your own LLM account and API keys (Anthropic, OpenAI, Bedrock, etc.) with OpenHands Open Source or OpenHands Cloud. | Separates platform cost from model cost and avoids lock-in; pay-as-you-go credits are only used when you opt into OpenHands’ at-cost provider. |
How It Works (Step-by-Step)
At a high level, OpenHands Cloud gives you two ways to pay for LLM usage: use the OpenHands at-cost LLM provider on pay-as-you-go credits, or bring your own keys and pay your LLM vendor directly. Here’s how the credit side works.
-
Choose the OpenHands at-cost LLM provider (per-project or per-workspace):
In OpenHands Cloud, you configure which model provider an agent uses. When you select the OpenHands at-cost LLM provider for a workspace, project, or specific task surface (Web GUI, CLI, SDK), that workload’s LLM calls will draw from your OpenHands Cloud pay-as-you-go credits. -
Run agents that consume tokens (and therefore credits):
As agents generate diffs, review PRs, write tests, or generate docs, they call the chosen LLM. With the at-cost provider, each prompt+completion uses tokens. Those tokens are measured and converted into a monetary charge. That charge is deducted from your pay-as-you-go credit balance in near real time. BYOK workloads do not consume credits; they consume your external LLM quota instead. -
Monitor usage and top up or adjust model strategy:
From your OpenHands Cloud account, you monitor LLM usage and remaining credit balance. When you get close to depletion, you either top up credits or adjust your configuration—for instance, routing long-running, high-token tasks to your BYOK model while keeping shorter interactive tasks on the OpenHands at-cost provider. As you scale agents across repos, tickets, and CI, this gives a clear lever for spend management.
Common Mistakes to Avoid
-
Assuming all LLM calls consume OpenHands pay-as-you-go credits:
Credits are only consumed when you use the OpenHands at-cost LLM provider. If you’ve configured BYOK, those calls are billed by your model vendor, not by OpenHands Cloud credits. Always verify which provider is selected for each workspace or agent class. -
Treating credits as “per user” instead of “per token”:
Credits aren’t tied to the number of seats or named users; they’re tied to LLM token usage. A single power user running continuous repo-wide refactors can consume more credits than a team of light, ad-hoc users. Plan budgets around workloads (PR volume, test runs, maintenance jobs), not just headcount.
Real-World Example
Imagine a small platform team adopting the OpenHands Cloud Growth Plan. They want fast time-to-value, so they start with the OpenHands at-cost LLM provider and a modest pool of pay-as-you-go credits.
- Day 1–7: They pipe GitHub PRs into OpenHands via the Web GUI and Terminal. Agents summarize PRs, apply review feedback, and generate tests. Each of these runs consumes tokens and chips away at their credit balance.
- Week 2: They add scheduled, headless runs via the SDK and CI to handle repo-wide dependency upgrades every night. Those runs are more token-heavy, so they see a noticeable jump in credit consumption.
- Week 3: To keep costs predictable, they send routine PR summarization and test fixes through the OpenHands at-cost provider (still on pay-as-you-go credits), but switch the heavy nightly upgrade jobs to their own Anthropic key (BYOK). From that point, the credits only track the interactive, collaborative work, while bulk automation usage is billed directly by their LLM vendor.
They’ve effectively separated two dimensions: platform orchestration (OpenHands Cloud) and LLM spend, with credits only attached to the at-cost provider where they want frictionless, managed usage.
Pro Tip: Treat credits like “fuel for specific lanes” rather than a single pool for everything. Keep the OpenHands at-cost provider enabled for workflows where convenience matters—ad-hoc PR review, production triage, or on-call fixes—and route long-running or experimental workloads to BYOK models where you already have volume discounts.
Summary
OpenHands Cloud pay-as-you-go credits are a straightforward, token-based way to use the OpenHands at-cost LLM provider without committing to fixed bundles or opaque tiers. When an agent uses the at-cost provider, its token usage is measured and charged against your credits; when an agent uses BYOK, your credits are untouched and your external LLM account is billed instead. This split lets you choose the right payment path per workload, scale agents from a single repo to thousands of runs, and keep both autonomy and cost under explicit control.