Agents that do the work and show it – kern
Unified session memory across Terminal, Slack, and Telegram beats fresh sessions per channel.
Agents that do the work and show it.
Agents serve live dashboards instead of forcing you to read chat logs.
AI developers and automation engineers
OpenHands · CrewAI · LangGraph
Unified session memory across Terminal, Slack, and Telegram beats fresh sessions per channel.
Agents build their own dashboards instead of chat—that's genuinely different.
Turns every agent client into a unified work log via a tiny MCP server and a single log_work call. Schema profiles and multiple sinks (jsonl, webhook, Postgres) let teams standardize payloads once and collect logs from Codex, ChatGPT, Claude, OpenClaw or cron jobs. Practical and low-friction — useful as an ops-level glue — but the dashboard is an in-repo MVP and delivery semantics (retries/schema evolution) will determine if it scales beyond a small team.
Folder-based agent structure with git-tracked plain text files is genuinely portable.
A folder is the agent—no server, no DB, just plain text files you can git-track.
Folder-based agent memory means zero infrastructure, just git-tracked plain text.