Skip to content

Supported Runtimes

CCCC can run multiple agent runtimes in the same working group. Each actor chooses one runtime, while the daemon keeps messaging, delivery tracking, tasks, context, and Web/IM control in one shared CCCC group.

Use cccc runtime list --all to see the full supported list on your machine, and cccc doctor to check which CLI runtimes are installed.

First-Class Runtimes

RuntimeRuntime idEntrypoint / surfaceMCP setup
Claude CodeclaudeclaudeAuto
Codex CLIcodexcodexAuto
GitHub Copilot CLIcopilotcopilotAuto
Cursor CLIcursorcursor-agentPrompt-assisted
Devin CLIdevindevinAuto
Kiro CLIkirokiro-cliAuto
Kilo Code CLIkilokiloPrompt-assisted
Antigravity CLIantigravityagyPrompt-assisted
Droid CLIdroiddroidAuto
AmpampampAuto
Auggie (Augment)auggieauggieAuto
Grok BuildgrokgrokAuto
Hermes AgenthermeshermesAuto through the user's Hermes profile
Kimi CLIkimikimiAuto
OpenCodeopencodeopencodeAuto via launch config
ChatGPT Web Modelweb_modelBound ChatGPT Web conversationBrowser delivery + remote MCP connector

custom is also supported as a manual fallback for any command-line agent that can be launched by CCCC.

Autonomy Defaults

CCCC applies conservative runtime-specific launch defaults for actors it starts. These defaults are intended to keep agent sessions moving without repeated approval prompts, while still leaving actor/profile commands editable in the Web settings.

Runtime idDefault commandPermission / autonomy behavior
claudeclaude --dangerously-skip-permissionsSkips Claude Code permission prompts.
codexcodex -c shell_environment_policy.inherit=all --dangerously-bypass-approvals-and-sandbox --searchBypasses Codex approvals/sandbox and preserves actor environment inheritance for MCP subprocesses.
copilotcopilot --allow-allAllows Copilot CLI tool execution without per-action approval.
cursorcursor-agent --yolo --approve-mcpsUses Cursor YOLO mode and approves MCP usage.
devindevin --permission-mode dangerousUses Devin's dangerous permission mode.
kirokiro-cli chat --trust-all-toolsTrusts Kiro tools for the session.
antigravityagy --dangerously-skip-permissionsSkips Antigravity tool permission prompts.
droiddroid --auto highStarts Droid in high-autonomy mode.
grokgrok --always-approveStarts Grok Build with approval prompts bypassed.
hermeshermes --tui --yoloStarts Hermes in TUI YOLO mode.
kimikimi --yoloStarts Kimi in YOLO mode.
opencodeopencode --autoAuto-approves OpenCode permission requests that are not explicitly denied.
ampampNo extra CCCC launch flag; Amp's current CLI default is already direct tool execution.
auggieauggieUse Auggie permissions or settings for per-tool approval policy; CCCC does not inject a broad wildcard permission rule.
kilokiloUse Kilo's kilo.jsonc permission settings or Auto Approve UI for broad approval policy.
web_modelN/ABrowser-delivered runtime; local CLI launch flags do not apply.
customUser commandCCCC preserves the user-provided command exactly.

Setup Commands

Most CLI runtimes can be prepared with cccc setup --runtime <id>:

bash
cccc setup --runtime claude
cccc setup --runtime codex
cccc setup --runtime copilot
cccc setup --runtime devin
cccc setup --runtime kiro
cccc setup --runtime droid
cccc setup --runtime amp
cccc setup --runtime auggie
cccc setup --runtime grok
cccc setup --runtime hermes
cccc setup --runtime kimi
cccc setup --runtime opencode

Prompt-assisted runtimes print an idempotent setup prompt or contract that you run inside that runtime:

bash
cccc setup --runtime cursor
cccc setup --runtime kilo
cccc setup --runtime antigravity

For a custom runtime, provide the command when creating or editing the actor:

bash
cccc actor add worker --runtime custom --command "my-agent --with-flags"

Runner Modes

Actors normally run in one of two modes:

  • PTY: the runtime runs in an embedded terminal. This is the broadest compatibility mode.
  • Headless: CCCC manages structured runtime I/O without a terminal. This gives tighter delivery and streaming control where supported.

Claude Code and Codex CLI support both PTY and headless operation. Most other CLI runtimes use PTY. ChatGPT Web Model is fixed to browser delivery plus a remote MCP connector.

ChatGPT Web Model

web_model does not use cccc setup. Create the ChatGPT Web Model actor from the CCCC Web group, then finish sign-in, MCP URL setup, and conversation binding in Settings > ChatGPT Web Model.

This runtime works with MCP-capable GPT-5.x ChatGPT Web sessions. GPT-5.x Pro sessions are advisory-only for this integration because they do not expose third-party MCP connectors.

For details, see ChatGPT Web Model Runtime.

Choosing a Runtime

Use a mixed group when different agents are good at different roles:

  • Use a Claude Code or Codex actor as the foreman when you want strong local coding orchestration.
  • Add a second runtime as reviewer to diversify feedback.
  • Use ChatGPT Web Model when you want a browser-backed GPT-5.x actor with CCCC MCP access.
  • Use custom only when the runtime is not first-class yet or needs a special command.

Each actor can have its own runtime, command override, private environment, and runner mode. Runtime state stays in CCCC_HOME, not in your repository.

Verification and Troubleshooting

bash
cccc runtime list --all
cccc doctor

Common checks:

SymptomCheck
Runtime is listed but unavailableInstall the CLI and make sure the command is on PATH.
MCP tools are missing in the runtimeRun cccc setup --runtime <id> or follow the prompt-assisted setup instructions.
Custom actor will not startEnsure --command is set; CCCC cannot infer a command for custom.
Existing actor does not pick up setup changesRestart the actor after setup or profile changes.
ChatGPT Web Model cannot call CCCCConfirm the public HTTPS MCP URL, ChatGPT connector setup, and bound conversation.

The Web UI also exposes runtime detection and actor configuration from the add/edit actor dialogs.

Released under the Apache-2.0 License.