Nous Research Rebuilds Hermes Agent Pricing Around Always-On Cloud Hosting

Nous Research has changed how customers pay for Hermes Agent. Newly purchased paid plans now include an always-on cloud agent, while daily and weekly usage caps replace the previous monthly credit allowance. The change lowers the setup barrier for hosted agents and limits how much capacity users can consume in a single burst. Existing subscribers can opt in through the billing portal.

Credits move to shorter clocks

Under the previous model, each tier provided a monthly pool of credits that subscribers could spend at their preferred pace, subject to additional rate limits.

The shorter allowance windows spread capacity across each billing period. A subscriber can no longer spend an entire month’s allocation during the first week, and unused capacity cannot be banked for a later burst. The daily limit constrains individual jobs, while the weekly limit governs sustained use across several days.

Exact cap values, reset times, time zones, and overage behavior are not specified in the available change summary. Developers evaluating the plans should verify those terms in Portal, along with any model-specific rate limits.

Hosting joins the subscription

Hermes Cloud provisions a hosted Hermes Agent with a persistent workspace and dashboard. The instance remains reachable while the user’s local machine is offline, allowing the agent to accept messages, execute scheduled work, and retain its operating context between sessions.

Cloud deployment previously sat alongside the Portal subscription as a separately funded service. Users needed an active subscription or at least $2 in Nous credits to deploy an instance, and the instance accrued daily charges based on use. Inference and tool calls were historically metered in addition to hosting.

New paid plans include the cloud-agent deployment entitlement. Subscribers should confirm whether instance, inference, and tool charges count toward the new caps or draw from a separate balance before leaving workloads online continuously.

Persistence unlocks unattended work

Hermes Agent remains free and MIT-licensed, so developers can continue to run it on their own hardware, a virtual private server, serverless infrastructure, or a GPU cluster. Portal provides the managed layer, consolidating authentication and billing for model and tool providers such as Firecrawl, FAL, Krea, Browser Use, OpenAI audio, and Modal.

A hosted runtime gives Hermes a stable place to use its persistent features. The agent can create and revise skills, search earlier conversations, maintain a workspace, and run scheduled jobs without depending on an open laptop. Practical workloads include:

• Generating and delivering daily reports through built-in cron jobs.
• Running nightly backups, recurring data collection, or weekly audits.
• Continuing long-running tasks while the user communicates through Telegram.
• Retaining skills and conversation history across sessions.

The new allowance windows also change workload planning. Regular, moderate activity fits naturally within daily and weekly caps, while large batch jobs may encounter a daily ceiling even when the subscriber has unused weekly capacity.

Check the limits before migrating

Existing subscribers must choose to migrate, which gives them time to compare their current usage with the new plan terms. The following details can materially affect production workloads:

1. Cap values: Compare both limits with peak daily use and total weekly consumption.
2. Reset rules: Check the reset time, time zone, rollover policy, and behavior after reaching a cap.
3. Burst workloads: Weekly batch jobs may fit the old monthly pool better if they exceed the new daily allowance.
4. Cloud metering: Confirm how hosting, inference, and external tools consume credits under the revised structure.
5. Additional limits: Check whether concurrency, model, or provider rate limits still apply alongside the plan caps.
6. Migration terms: Verify whether switching is reversible and whether the subscription price changes.

The billing redesign aligns Portal more closely with persistent, hosted use. It simplifies cloud-agent deployment and makes recurring capacity easier to estimate, while reducing flexibility for subscribers whose workloads arrive in occasional large bursts.