Use cases

Pasture Protocol is not one chatbot feature. It is a local runtime that connects chat channels, skills, memory, agents, and dashboard workflows.

Everyday assistant scenarios

ScenarioHow Pasture handles it
Ask from WhatsApp or TelegramPasture receives the message, finds the routed agent for that chat, loads relevant skills, and replies back in the same channel.
Set remindersNatural language like 'tomorrow at 9' or 'every Monday' becomes a one-shot or recurring job tied to the same chat where you asked.
Search the webThe search skill fetches fresh results, summarizes them, and avoids pretending it knows live information from memory.
Browse a websiteThe browser skill opens a local Chromium tab, can click, fill, scroll, screenshot, and keep the tab for follow-up questions.
Understand images or voiceImages are passed to a vision-capable model or fallback. Voice notes are transcribed before the agent reasons over them.

Work and project scenarios

ScenarioHow Pasture handles it
Remember facts and decisionsMemory saves durable notes only when asked, then retrieves them semantically later by topic, date, or project context.
Work on a software projectThe project workflow checks the dashboard catalog, asks for missing setup details, proposes a mission, waits for approval, then tracks tasks.
Manage GitHub workWith a token, Pasture can read repos, issues, and PRs. Write actions like branches, comments, PRs, and merges require confirmation.
Handle Gmail and CalendarThe Google skill uses the local gog CLI, answers from returned data, and confirms before sending email or creating/deleting events.
Inspect serversThe SSH inspection skill runs read-only checks for disk, processes, logs, services, and Docker state. It is not enabled in groups.

Home, agents, and longer-running work

ScenarioHow Pasture handles it
Control home devicesHome Assistant commands map natural phrases to lights, switches, thermostats, scenes, scripts, and sensors.
Ask about rooms or devicesIf an entity ID is unknown, Pasture searches by friendly room/device name, then uses the matching entity for the action.
Use specialist agentsMain can delegate to linked agents such as writer, backend, reviewer, or assistant, then combine their answers into one reply.
Run persistent missionsGoals keep owner, status, plan, progress, evidence, next run time, and subgoals so work can continue beyond one chat turn.
Follow up after silenceTide can send one AI-composed follow-up after a configured quiet window, rather than repeatedly nagging.

How to decide where to read next

  • If you want the request lifecycle, read How requests are handled.
  • If you want WhatsApp, Telegram, DMs, and groups, read Channels.
  • If you want specialist personas and delegation, start with Agents.