01 / Gateway
Slack and email as work surfaces.
Hermes can take operational requests from Slack and deliver reports or automation outputs through email when that is the better channel.
A working AI operator setup around my own work system: Slack and email intake, Obsidian knowledge capture, local terminal/file execution, scheduled automations, and Cloudflare-backed webhooks.
Hermes is not just a chat window for me. It is an operator layer that can receive a request, inspect files, run commands, update notes, schedule work, and report back through the same channels I already use.
The point is practical implementation: the assistant has durable context, project-specific skills, local tools, and a human-readable note system, so useful outcomes do not disappear into chat history.
Hermes can take operational requests from Slack and deliver reports or automation outputs through email when that is the better channel.
Durable outcomes are promoted into Obsidian project and daily notes. The vault stays human-readable while session history and retrieval support recall behind the scenes.
The agent can read and edit files, run terminal commands, inspect repositories, work with git, and verify that changes actually landed instead of only describing a plan.
Stable preferences, environment details, and reusable procedures are captured so future sessions need less repeated steering and fewer rediscovered steps.
Cron-style Hermes jobs support recurring checks, one-shot reminders, and deliverable reports that come back to Slack or email with useful context.
Cloudflare DNS, Pages, tunnels, and webhook routes give the system a practical way to expose selected endpoints without broad inbound server access.
The system is already used for project notes, website updates, infrastructure follow-up, daily capture, deployment checks, and small operational tasks.
Work is grounded in files, commands, commits, notes, and explicit verification. The useful part is not blind autonomy; it is moving faster with visible state.
The same stack can support client-facing agent workflows: messaging gateway, local execution, knowledge base, scheduler, webhooks, and a deployable operating model.
Hermes is reachable from Slack for direct tasks and follow-ups, with platform delivery available for Slack and email outputs.
Project notes, daily notes, and work/personal separation keep durable outcomes accessible without treating raw transcripts as the note system.
The implementation now supports local execution, scheduled jobs, reusable skills, persistent preferences, and parallel investigations when a task is too bulky for one thread.
The public framing is simple: implemented an AI operator layer for knowledge capture, automation, and delivery across Slack, email, Obsidian, local tools, and Cloudflare-backed webhooks.
This is the kind of project I want to show publicly: applied AI infrastructure that is small enough to inspect, useful enough to operate daily, and concrete enough to rebuild for other workflows.
Back to projects