English | δΈζ
Terminal coding agent powered by open-source models β Rust-built, Claude Code compatible
DeepSeek-V4-Pro + Mimo-2.5Pro + GLM-5.1 driven, zero migration from .claude/ config, runs on RISC-V.
npm install -g @cc-claw/code
π Official Website: cc-claw.com
Why cc-code Β· Core Capabilities Β· Install Β· Nobody Coding Β· Acknowledgments
Click to collapse
Kimi K2.6 is an open-source, native multimodal agentic model from Moonshot AI, built for long-horizon coding, coding-driven design, and swarm-based task orchestration. It handles complex end-to-end engineering work across front-end, DevOps, performance optimization, and full-stack workflows. Register here
 |
Top-tier model MiMo V2.5 from Xiaomi. Register with invite code: both get Β₯10 API credit + 10% off first order. Invite code: JBEYTF. Register here (auto-filled on registration Β· credit valid for 40 days) |
 |
GLM Coding Plan from Zhipu AI β top-tier coding model in China, compatible with 20+ mainstream tools, best value. Join now |
| Comparison | Other Terminal Agents | cc-code |
|---|---|---|
| Runtime | Node.js / Bun, easily eats 1GB RAM | Rust native, fast startup, ~50MB memory |
| Model Lock-in | Locked to one LLM | Switch freely: Anthropic, OpenAI-compatible, DeepSeek, GLM |
| Prompt Cache | Recompute every turn, wasting tokens | Frozen system prompt, 95-99% cache hit rate |
| Tool Loading | All tools stuffed into every request | Core tools resident, rest lazy-loaded via Tool Search |
| IDE Integration | Terminal only | ACP protocol, Zed and other IDEs connect directly |
| Claude Code Ecosystem | Incompatible | Use .claude/ config, agents, skills, hooks, MCP directly |
| Capability | Description |
|---|---|
| Rust Native | Fast startup, low memory, zero runtime overhead |
| Context Optimized | System prompt frozen + dynamic content isolated, no token waste |
| Multi-LLM Support | Anthropic / OpenAI-compatible APIs, DeepSeek, GLM β switch freely |
| Claude Code Compatible | .claude/ config, agents, skills, hooks, MCP, sub-agents all reusable |
| Streaming Markdown | Code blocks, tables, diffs rendered in real-time |
| ACP Protocol | Connect to Zed and other IDEs, or build your own "Cloud Code" platform |
| Auto Compact | Long sessions auto-compressed, stays fast and cheap |
| Sub-Agent Concurrency | Background sub-agents run in parallel with fork and background modes |
| HITL Approval | Sensitive operations auto-intercepted with auto-classifier and shared-mode |
| Dual-Engine File Search | Grep/Glob prefer the external ripgrep binary, falling back to the built-in Rust engine |
| Nobody Coding | 99% of code produced by DeepSeek, Mimo, and GLM β humans decide what, AI figures out how |
| Feature | Version | Description |
|---|---|---|
| ACP Permission Forwarding | v0.6.94 | cc-code acp now forwards tool permission requests to the IDE client via session/request_permission instead of auto-approving (fail-closed: client unsupported / call failure / unknown option = denied); stdio default permission mode changed from Bypass to AutoMode (use session/set_mode for unattended bypass) |
| Fail-Closed Permissions | v0.6.90 | Approval is now on by default: unset YOLO_MODE no longer bypasses HITL (explicit -y/--yolo or YOLO_MODE=true to skip); HITL gate evaluates the post-rewrite command; git clone hardened against option injection |
| Three-Choice Tool Approval | v0.6.84 | HITL approval dialog now offers allow-once / allow-for-session / reject; choose "session" to stop repeated prompts for the same (tool, path) within the session (path-level, session-scoped approval memory) |
| Spinner Color & Stuck Detection Fixes | v0.6.83 | thought for Ns state word now always muted (only in-progress states warm up with time); stuck-detection no longer misfires on blank thinking fingerprints; stuck-detection switch-strategy hint switched to English for consistency |
| Thinking Status Line & Tool Summary | v0.6.82 | Spinner third field 4-state machine (thinking / thought for Ns / still thinking / thinking more); message-area thought line β Thought for Ns, <action counts>; consecutive thinking+read-only-tool rounds merged into one line; Bash non-verbose output summary with ... (N more lines) (ctrl+o to expand) |
| Semantic Gate Hardening & Two Permission Modes | v0.6.81 | HITL gate rebuilt on upstream Jev semantics: curl|bash and interpreter-family bypasses are now hard-denied, write path traversal can no longer skip the gate, and judge unavailability returns to human confirmation instead of blocking the whole session; permission modes collapsed to auto/bypass (default auto), with the deterministic layer no longer sharing a switch with the semantic judge |
| Session Recap Persistence | v0.6.80 | Recap and task-summary lines persist to ThreadMeta (latest_recap/last_task_summary), no longer lost on restart |
| /recap Command & Auto Recap | v0.6.76 | /recap (/away /catchup) summarizes goalβtasksβnext step via aux_model; auto-recap on terminal blur + β₯3 turns + 60s idle (/config toggle, PERI_AUTO_RECAP_* env); non-streaming Anthropic responses adapt to reverse-proxy OpenAI format |
| Tool Schema Validation & Input Pump | v0.6.75 | Human-readable schema errors aligned with Claude Code's formatZodValidationError + tool-mismatch heuristics (suggest_tool_mismatch) + failure circuit breaker; dedicated InputPump safely enables mouse hover and fixes scrollbar drag |
| RTK Output Filtering | v0.6.74 | Filters noisy RTK git status (clean β nothing to commit) and drops toxic generic folding that swallowed code context |
| Windows Console Isolation & Scroll Debounce | v0.6.73 | CREATE_NO_WINDOW isolates child consoles, eliminating full-screen flicker from PHP etc.; wheel debounce and smooth scrollbar thumb drag on long content |
| Clickable Markdown Links | v0.6.72 | Markdown hyperlinks in the message area open the default browser cross-platform |
Older releases (v0.6.0 β v0.6.71) are listed in the CHANGELOG.
Binaries available for macOS (x86_64 / Apple Silicon), Linux (x86_64 / aarch64 / riscv64), and Windows (x86_64).
npm install -g @cc-claw/codenpm update -g @cc-claw/codecurl -fsSL https://raw.githubusercontent.com/cc-claws/cc-code/main/scripts/install.sh | bashirm https://raw.githubusercontent.com/cc-claws/cc-code/main/scripts/install.ps1 | iexNobody Coding means exactly what it sounds like. No human wrote a single line of cc-code β not the architecture, not the TUI, not the harness tuning that makes open-source models reliable in an Agent loop. Humans decide what. AI figures out how. You're not pair programming β you're product managing an engineer that never sleeps. 99% of cc-code was built this way.
Recent commits are almost entirely DeepSeek, Mimo, and GLM. Claude was just there in the beginning.
| When you... | Pipeline kicks off |
|---|---|
| Find a bug or piece of tech debt | issue-create β systematic-debugging β writing-plans β subagent-driven-development β issue-archive β improve CLAUDE.md |
| Want to build a new feature | grill-me β writing-plans β subagent-driven-development |
| Notice the codebase getting messy | slop-cleaner β improve-codebase-architecture β writing-plans β subagent-driven-development |
| Need someone to grok the architecture | teacher β assign a task β teacher |
cc-code/
βββ cc-agent/ # Core: Agent loop, tool system, persistence, telemetry
β βββ README.md # Agent framework guide
βββ cc-middlewares/ # Middleware: filesystem, terminal, MCP, Hooks, etc.
β βββ README.md # Middleware overview
β βββ CLAUDE.md # Development guide and traps
βββ cc-tui/ # TUI application (Ratatui)
β βββ README.md # TUI usage guide
β βββ CLAUDE.md # Development guide and traps
βββ cc-acp/ # ACP service layer: bridges TUI/IDE with Agent
β βββ README.md # ACP architecture and data flow
βββ cc-widgets/ # Widget component library
β βββ README.md # Component list and examples
βββ cc-lsp/ # LSP client library
β βββ README.md # LSP operations and config
βββ langfuse-client/ # Langfuse telemetry client
β βββ README.md # Telemetry config and usage
βββ npm/ # npm package: postinstall script + shell wrapper
β βββ README.md # npm package docs
βββ scripts/
β βββ install.sh # macOS / Linux installer
β βββ install.ps1 # Windows installer
βββ side-projects/ # Experimental projects (llm-gateway, etc.)
βββ spec/ # Design docs and specs
β βββ global/ # Global architecture docs
β βββ issues/ # Issue analysis docs
β βββ prd/ # Product requirement docs
βββ CLAUDE.md # Project development guide
βββ CHANGELOG.md # Version changelog
βββ CONTRIBUTING.md # Contribution guide
βββ README.md
βββ README_ZH.md
βββ LICENSE # Apache 2.0
| Project | Description |
|---|---|
| Peri (KonghaYao) | This project is forked from Peri, distributed under Apache 2.0. Credit to the original author. |
| Superpowers & Matt Pocock's Skills | Skill suites driving cc-code's AI engineering workflow |
| ACP | Open protocol for agent-IDE communication |
| rmcp | Rust MCP client library |
| Ratatui & Tokio | TUI framework and async runtime |
| Langfuse | LLM observability |
| Zed | First ACP-compatible IDE, proved the protocol works |
Apache License 2.0 β free to use, modify, and distribute, including commercial use.