Work from the repository.
- Your selected model and instructions
- Selected skills and child agents
- Repository files and local context
For example: plan a refactor from the code and tests already in the repo.
Agent Farm / Field notes 01
A planner needs different instructions from an implementer. Save those choices once, then take the right setup into your next repository.
The toolkit
Choose a saved agent with its own model, instructions, and skills.
Let init check prerequisites and install your starting profiles and skills.
Bring a project’s remote or local MCP tools into a launch.
Load a profile’s skills into your native user directory across repositories.
Share versioned profiles, agents, and skills; detect conflicts with local edits.
Use headless execution with an initial message for scripted work.
Choose native delegation or a separately launched process for each child.
Launch Claude Code or Codex with the configuration your agent selects.
Preview a profile’s settings, connections, and source files.
Work alongside the repo’s own AGENTS.md, CLAUDE.md, and references.
From setup to session
After agent-farm init, launch the interactive picker with agent-farm. This illustrated sequence shows the path to a session.
$ agent-farm › planner implementer reviewer
Choose the worker and playbooks for this job.
Project tools None › my-project MCP connections included
Choose project connections when you need them.
Working directory ~/repos/my-app Profile + tools + repo ✓
Choose where the agent will read and work on files.
.agent-farm/generated/ instructions · skills · tools Claude Code or Codex →
Agent Farm builds the bundle and opens the selected native CLI.
Illustrative screens. Your installed profiles and workspaces determine the choices shown.
agent-farm run planner --workspace my-projectReview the launch permissions before your first run.
The same worker, more context
A repository tells the agent where to work. A workspace supplies named connections to external tools, independently of that folder.
For example: plan a refactor from the code and tests already in the repo.
For example: use a connected tracker and error service to investigate a reported bug.
Add your provider’s MCP endpoints and authenticate them first. Agent-defined connections and native global tools can also apply.
Build a workspace configuration