Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/cline/kanban/llms.txt

Use this file to discover all available pages before exploring further.

Kanban works with any CLI coding agent you have installed. On startup, it scans your PATH for the binary of each known agent. Every agent it finds becomes available as a task runner on your board — no additional configuration required.

Agent detection

Kanban checks for each agent’s binary name in your PATH. If the binary is present, the agent is listed as available in the board UI. If none are found, Kanban prompts you to install one.

Agents

Binary: claudeClaude Code is Anthropic’s official CLI coding agent.Install it by following the Claude Code quickstart.When Kanban runs a task with Claude Code in autonomous mode, it passes --dangerously-skip-permissions to suppress interactive permission prompts.
Binary: codexOpenAI Codex is OpenAI’s CLI coding agent.Install it from github.com/openai/codex.When Kanban runs a task with OpenAI Codex in autonomous mode, it passes --dangerously-bypass-approvals-and-sandbox.
Binary: clineCline is the open-source CLI coding agent from Cline Bot Inc.Install it from github.com/cline/cline.When Kanban runs a task with Cline in autonomous mode, it passes --auto-approve-all.
Binary: opencodeOpenCode is an open-source terminal coding agent.Install it from github.com/sst/opencode.OpenCode has no autonomous-mode flags at this time. Full launch support is coming in a future release.
Binary: droidDroid CLI is Factory AI’s command-line coding agent.Install it by following the Droid CLI quickstart.When Kanban runs a task with Droid CLI in autonomous mode, it passes --auto high.Full launch support is coming in a future release.
Binary: geminiGemini CLI is Google’s open-source coding agent for the terminal.Install it from github.com/google-gemini/gemini-cli.When Kanban runs a task with Gemini CLI in autonomous mode, it passes --yolo.Full launch support is coming in a future release.

Autonomous mode

When you hit the play button on a task card, Kanban launches the agent with additional flags that bypass interactive approval prompts. This is what allows agents to complete work end-to-end without human intervention at every step.
Autonomous mode uses experimental flags that allow agents to take actions without asking for confirmation. Review the changes in each worktree before committing.
The flags passed per agent are:
AgentAutonomous flag(s)
Claude Code--dangerously-skip-permissions
OpenAI Codex--dangerously-bypass-approvals-and-sandbox
Cline--auto-approve-all
Droid CLI--auto high
Gemini CLI--yolo
OpenCode(none)

Full launch support

Full automated launch support — where Kanban creates a worktree, starts the agent, monitors it via hooks, and handles commit or PR creation — is currently available for:
  • Claude Code
  • OpenAI Codex
  • Cline
OpenCode, Droid CLI, and Gemini CLI are recognized and can be detected, but full launch support for these agents is still in progress.