Agent team
Named agents with live context, task counts, inboxes, and delegation state.
pastureprotocol.com
Completely free to install and use
Build agent workflows that classify, route, delegate, remember, and act across tools without burying behavior inside application code.
Install in one line
curl -fsSL https://raw.githubusercontent.com/bishwashere/pastureprotocol/master/install.sh | bashiwr -useb https://raw.githubusercontent.com/bishwashere/pastureprotocol/master/install.ps1 | iexThen run pasture start and pasture dashboard
Major features
Pasture is built around a dashboard that shows what the agent system knows, who is working, what is running, and which tools are available.
Named agents with live context, task counts, inboxes, and delegation state.
Memory, imported context, chat history, and recurring terms that shape recall.
Project work becomes missions, task lanes, recent movement, and attention queues.
Review work by status, assignee, labels, source, timeline, and completion window.
Track project setup, connectors, branches, update chains, and long-running context.
Open persistent missions, inspect progress, and manage linked task trees.
Review throughput, delegation counts, execution time, and skill usage across agents.
What can you do?
The useful part is not one command. It is the loop: chat, context, tool use, memory, delegation, and a dashboard that keeps the work visible.
Work on NextPostAI. What should we do next?
Pasture checks the project catalog, asks for missing setup details, proposes a mission and tasks, waits for approval, then tracks progress in Projects, Missions, and Tasks.
What did we decide yesterday about pricing?
The agent searches chat history, long-term notes, and semantic memory, then answers with the relevant context instead of starting from scratch.
Draft the release notes and ask the backend agent to verify the PR.
The main agent can delegate bounded work to specialist agents, collect their results, and keep the whole thread visible in the team activity feed.
Every weekday at 8pm remind me to log my hours.
Pasture turns natural language into a local recurring cron, stores it in the reminder list, and sends the reminder back through your chosen channel.
Open the site, inspect the pricing page, and summarize what is confusing.
With skills enabled, an agent can browse pages, search the web, read files, inspect images, query GitHub, or work with connected services.
Keep checking whether this task is blocked and tell me what needs my input.
Missions, tasks, activity, reports, and attention queues show what moved, what finished, and what needs a human decision.