English · 简体中文
Proactive Agent keeps project progress up to date across your AI conversations and suggests useful next steps that you can accept or ignore.
Supports Hermes and OpenClaw.
From connected conversations, the service organizes each project's goals, latest progress, blockers, and next steps. The dashboard brings projects together in one place; expand a project to see its tasks and progress bars. As new progress is recorded, task statuses and their progress bars update too.
Useful follow-ups appear as short suggestion cards above your projects. Accept a suggestion to have the assistant continue the action in the original connected conversation, or ignore it without running the proposed action.
Demo screenshots use sample data; the interface shown is in Chinese.
Materials from previous work may still be useful for a new task. The service can suggest relevant templates, reports, or conclusions you have already shared. For example, when preparing a quarterly presentation, it can suggest reusing the previous quarter's slide structure and adding the latest figures, reducing the work of starting over.
New findings may add to or change an earlier conclusion. The service can connect them with related work and suggest an update. For example, a new retention analysis may be useful not only for the current question, but also for revising a weekly report written earlier.
When several projects are moving at once, the service brings their key progress and next steps into a short daily recap. It opens on your first dashboard visit of the day, with no chat request needed. Dismiss it when you're done, or reopen it from the daily-report button.
Follow the setup Skill to download and install the verified runtime package from GitHub Releases.
Use the setup Skill with your agent to install the service and connect your assistant, or follow the installation guide.
After setup, start the service and leave it running:
sn-proactive-agent serve --web-onlyFor a connected Hermes installation, open it in another terminal:
hermes --tui --accept-hooksFor your first visit, open the private authorization link printed by the service. After authorization, use the dashboard to view suggestions, project progress, and your daily recap. To get a new link, run this in your own terminal; do not share its output:
sn-proactive-agent dashboard --url http://127.0.0.1:8080Upgrading from v0.1.4 or earlier? Follow the upgrade steps to update both the runtime and your Connector. Updating the Python package alone is not enough.
Records are stored locally; relevant conversation and project content may be sent to your configured model service for AI processing.


