CCB v6(Linux) - Infinite Parallel Agents Edition

Native multi-agent runtime for terminal split panes Claude · Codex · Gemini · OpenCode · Droid Visible concurrency, native communication, project-scoped runtime

English | Chinese

Introduction: CCB v6 is the infinite parallel agents edition. It turns split-pane collaboration into a native multi-agent runtime where agents can run side by side, hold independent roles and personalities, and delegate to each other through a stable built-in communication layer.

Why Parallel Agents Matter

Parallel agents are not just "more panes on screen". In CCB, each agent can own a fully independent role, task stream, skill library, and personality.

CCB provides the runtime foundation for stable agent-to-agent communication and effectively unbounded delegation. It supports arbitrary agent naming and window arrangement, per-agent control, broadcast dispatch, and point-to-point communication.

🚀 User Commands

• ccb start, restore, and attach CCB in the project directory from the terminal
• ccb -s safe mode
• ccb -n rebuild the project .ccb state
• ccb kill close CCB
• ccb kill -f deep-clean exit

💬 Communication Usage

Inside a provider / agent runtime:

• /ask all "sync on the latest repo state" broadcasts one message to all live agents.
• /ask reviewer "review the new parser change" sends work to the named [reviewer] agent.

Typical pattern:

• use ask all for one-shot broadcast or global sync
• use ask agent_name for targeted delegation
• use implicit skill-based calls; agents invoke ask themselves, and skills are currently auto-installed into Codex, Claude, and other providers
• if you only occasionally need to inspect replies manually, pend and watch are secondary tools

🛠 Config Control

ccb is controlled by .ccb/ccb.config. That file defines agent names, pane layout, and whether an agent runs inplace or in a separate git worktree.

Quick rules:

• agent_name:provider defines one agent. agent_name is also the pane label and logical runtime name.
• cmd adds one shell pane.
• ; splits panes horizontally from left to right.
• , splits panes vertically from top to bottom.
• Default workspace mode is inplace. If one agent needs an isolated git worktree to avoid conflicts, write agent_name:provider(worktree).

Example:

cmd; writer:codex, reviewer:claude; qa:gemini(worktree)

This layout means:

• left pane: cmd
• right side: a vertical stack
• top-right pane: writer
• bottom-right side: a horizontal split between reviewer and qa
• qa runs in an isolated git worktree; writer and reviewer run inplace in the main project

Historical note: older release notes below may mention askd, legacy flags, or removed commands. Those references are kept only as changelog history and do not redefine the current CLI surface.

• CodeIsland Hook Assets Inherited: managed Claude startup now copies referenced source-home hook assets such as .codeisland/ when inherited Claude hooks call $HOME/.codeisland/..., preventing missing-hook failures inside isolated Claude homes
• Config Boundary Preserved: third-party hook assets are copied only when Claude config inheritance is enabled and the inherited hook payload actually references that home-relative asset path
• Diagnostics Redaction Extended: diagnostic bundles now exclude copied .codeisland/ provider-state assets while still including ordinary managed Claude settings for support

• Claude Official Login Source Home Fixed: managed Claude startup now treats .ccb/agents/*/provider-state/*/home as an isolated runtime home, not the user's source home, so official browser-login credentials are copied from the real account home
• Claude Credential Path Coverage: managed Claude homes now project Claude Code official-login credentials from .claude/.credentials.json while retaining compatibility with .config/claude-code/auth.json
• Regression Coverage Added: tests now lock source-home fallback, launcher projection, diagnostics redaction, and workspace preparation for official Claude login inheritance

• Claude Official Login Projection: managed Claude homes now project Claude Code official login credentials from .claude/.credentials.json, so browser-login-backed auth can be inherited into isolated CCB runtimes instead of only API-token-based settings auth
• Managed Login Auth Retention: when global Claude auth artifacts disappear but managed Claude state already holds a valid project-scoped login, startup now preserves that managed login auth across restart instead of silently dropping it
• Auth Cleanup And Regression Coverage: disabling auth inheritance now clears stale copied Claude login credentials, and targeted tests now lock the projection, cleanup, and launcher startup paths

• Empty Gemini Hook Replies No Longer Burn Jobs: managed Gemini AfterAgent hooks that fire with an empty reply now downgrade to incomplete instead of terminalizing as a false exact completion
• Exact Hook Polling Becomes Safer: Gemini exact-hook polling now ignores completed hook artifacts with no reply text, allowing observed session-stability or timeout reliability paths to converge the request instead of accepting a blank terminal result
• Regression Coverage Added: targeted tests now lock the empty-reply guard at both the finish-hook artifact writer and Gemini execution-service polling layers

• Gemini Endpoint Override Restored: managed Gemini startup now preserves GOOGLE_GEMINI_BASE_URL end to end, so custom endpoint and proxy-backed Gemini CLI setups no longer fall back to Google's default production API host
• Gemini Model Env Allowlisted: control-plane and provider-profile env filtering now preserve GEMINI_MODEL, allowing isolated Gemini agents to keep explicit model selection instead of silently dropping it at startup
• Config Shortcut Alignment: Gemini key / url shortcuts now materialize the same environment variables the current Gemini CLI actually reads, keeping explicit config-based routes aligned with shell-level env behavior

• Codex Plugin Projection Fixed: managed Codex homes now project plugin-bundle authority under .tmp/plugins/ and .tmp/plugins.sha, so isolated agents inherit the marketplace catalog and installed plugin assets they actually need instead of starting with plugin-enabled config but missing bundles
• Plugin Refresh Semantics Tightened: startup now refreshes managed plugin projections as one authority unit, removes stale managed plugin residue when the source projection disappears, and keeps a cheap no-recopy fast path when the source plugin freshness marker is unchanged
• Cmd Shell / Session Env Hardening: the cmd pane now directly execs the resolved user shell and preserves ordinary user-session transport variables such as DISPLAY, WAYLAND_DISPLAY, DBUS_SESSION_BUS_ADDRESS, XAUTHORITY, and SSH_AUTH_SOCK, improving fish/zsh and GUI-command compatibility

• Codex Explicit Route Authority: managed Codex homes now materialize agent-local config.toml and auth.json as the sole authority for explicit key / url routes, so agent-scoped API overrides replace inherited global provider routes instead of drifting back to system config
• Codex Session Namespace Rotation: managed Codex startup now fingerprints explicit route authority, stamps reusable session bindings with that authority, and rotates stale sessions/ namespaces before launch when the bound route no longer matches
• Foreground Attach UX Hardening: interactive ccb startup now seeds tmux namespace creation from the real terminal viewport and issues a best-effort client refresh after attach so first paint matches the current terminal size without manual redraw

• Managed Claude Auth Preservation: managed Claude homes now preserve agent-local login auth when the global Claude home has been logged out, so a project-scoped re-login survives restart instead of re-entering a browser-link loop
• Auth Projection Semantics Tightened: Claude startup still refreshes source auth when it exists, but no longer treats missing source auth as an instruction to blank managed auth; disabled auth inheritance still clears stale copied auth state
• Startup Regression Coverage Expanded: targeted regressions