A running log of what's shipping in Hermes Agent — merged pull requests from NousResearch/hermes-agent, browsable one day at a time with plain-English AI summaries of what changed. Checked every 30 minutes.
Last checked 12 Sept 2026, 05:24 UTC
You can now have natural, back-and-forth voice conversations with Hermes using a new "GPT-Live" mode. When you speak, Hermes understands your requests and uses its full capabilities to respond, then speaks the answer back to you.
The app now provides a smoother experience when connecting new services by showing a clear "Connected" status after sign-in. This change also improves the initial setup process, making it more reliable and easier to follow.
When you first set up Hermes and tell it which apps you use, it will now connect those apps before starting your first task. This means Hermes can immediately use your connected services to help you.
This update clarifies the internal instructions and explanations within the app's setup process. This makes it easier for developers to understand and maintain the code, ensuring a smoother experience for users during their first launch.
Hermes now automatically serves every profile found in the `profiles/` directory, simplifying management by removing the need to explicitly list which profiles to run. This change ensures that each profile correctly uses its own specific settings and environment variables, preventing conflicts and improving the isolation of individual profile operations.
The dashboard and `gateway stop` command now correctly recognize and manage profiles served by the multiplexer as running through it. This ensures a consistent experience when checking profile status or stopping services, preventing confusion where a profile was visibly running but reported as stopped.
Hermes Agent now correctly applies each profile's individual settings when running adapters within a multiplexed gateway. This ensures that features like mention requirements, reaction settings, and listener ports function as expected for each profile, just like when they run on their own.
When running multiple agent profiles at once, each profile now correctly uses its own specific settings for things like the current working directory, approved tools, secret redaction, and tool policies. This ensures that each profile behaves independently and as expected, rather than inheriting settings from other profiles or the main process.
Now when you reply to a WhatsApp message that contains an image, that image will be attached to your reply. This makes conversations clearer by ensuring the agent sees the full context of your message, even if it's a picture you or the bot sent earlier.
Hermes Agent can now properly recognize repositories you've already marked as safe to use, preventing errors when working with shared or containerized code. This ensures your Git operations in isolated environments will succeed without unexpected ownership issues.
Plugins can now track a Group Chat member's actions in real-time during a turn, providing greater visibility and control over their behavior. This update allows plugins to see details like tool usage and message updates as they happen, enabling more responsive and sophisticated automation.
The project's automatic formatting tool has been updated to better clean up code style issues. This ensures the code remains consistent and easier for everyone to read and contribute to.
Previously, when running multiple Hermes profiles at once, certain features like scheduled tasks and notifications might not have worked exactly as they would if each profile was run individually. This update ensures that all features within a multiplexed profile now behave identically to how they would when that profile is run on its own, improving consistency and reliability.
The project's code formatting has been automatically improved to ensure consistency and readability, making it easier to understand and maintain. This update helps keep the codebase clean and well-organized without requiring manual intervention.
Bots can now use pets that you've hatched yourself as their avatar. This fixes an issue where these custom pets wouldn't load or display correctly.
The code formatting has been automatically improved to ensure consistency and readability. This makes it easier for developers to understand and contribute to the project.
Code formatting and style issues in the JavaScript parts of Hermes Agent are now automatically fixed. This change helps keep the project's code clean and consistent without manual effort.
The project's code formatting has been automatically cleaned up to ensure consistency and improve readability. This makes it easier for developers to understand and contribute to the project.