Docs

Keshro Documentation

Keshro adds planning and execution coordination to your existing AI agent. Install it, log in, and load Keshro in your agent session to get started. In Claude Code, that can be /keshro.

Agent Mode

Install

Agent Mode still relies on the Keshro CLI being installed locally so the /keshro skill can authenticate, create plans, and drive execution from inside your coding agent.

uv tool install keshro

Then run keshro login to connect your account and install the local Claude Code and Codex integrations.

Load KeshropreferredAgent Mode

Load Keshro in your agent session. Your agent scans the codebase, asks clarifying questions, generates the plan, and starts executing.

Setup: keshro login to authenticate and install local Claude Code and Codex integrations.

Works with: Claude Code (load /keshro), Codex (load the Keshro skill), Cursor (keshro setup for repo rules)

After installing, the recommended flow is to load Keshro in your agent session. In Claude Code, that can be /keshro, but the same workflow is available directly in the CLI.

Prefer the agent flow for normal use. If you want direct control, every step is also available through the CLI.

"Plan a migration from one platform to another"

Keshro detects the migration, scans your repo, asks clarifying questions, and generates a task list with dependencies.

"Use keshro to plan this project"

Keshro generates a dependency-ordered execution plan from your description or imported issues.

"Execute the plan" or "Run it"

Your agent starts working through tasks, in parallel across isolated worktrees by default, on whichever executor each task is configured for.

"Run this on Managed Agents" / "Use the hosted runtime for this task"

Asks the agent to switch the task's executor to Anthropic Claude Managed Agents before executing, so the task runs in a hosted cloud container instead of a local subprocess. Requires Managed Agents config in your Account or Org settings.

"What's the status?"

Shows progress, completed tasks, and what's next. Tasks running on Managed Agents are tagged so you can tell which runtime each one used.