OrbCode CLI

​

Why OrbCode CLI

• Agentic in the terminal — read, edit, search, and run commands across your whole workspace without leaving the shell.
• Same Axon models as Orbital — axon-eido-3-code-pro, axon-eido-3-code-mini, axon-code-2-5-pro, and axon-code-2-5-mini, with native JSON tool calls and image input.
• Multi-provider — route to Anthropic Claude models or any OpenAI-compatible endpoint via the Vercel AI SDK, reusing the same agent loop, tools, and approvals.
• MCP servers — connect external tools through the Model Context Protocol; servers appear as mcp__<server>__<tool> and can be called by the model.
• Skills & AGENTS.md memory — load reusable instruction sets on demand, and inject project/user instructions into every system prompt. The repo-level file lives at .orb/AGENTS.md, the same folder the Orbital IDE reads.
• Linked repositories (/link) — point this repo at other repos on your machine so a change here is checked for impact on (or propagated to) them. Links live in the shared .orb/links.json.
• /init — analyze the codebase and write a concise .orb/AGENTS.md for cold-start, ~150 lines covering project structure, architecture, business-logic mapping, and code patterns/conventions.
• Streaming TUI — markdown rendering, live reasoning under a collapsible “Thinking” block, and per-tool activity rows with diffs.
• Approvals & safety — read-only tools run freely; file edits and commands prompt first, and dangerous commands can never be auto-approved.
• Headless mode — pipe a single prompt with -p for scripts and CI.
• Claude-Code-compatible hooks — block, auto-approve, rewrite, or inject context at fixed points in the agent loop.

​

Quick start

npm install -g @matterailab/orbcode

orbcode

​

Models

​

Other providers

• provider: "anthropic" → native /v1/messages. Built-in Claude models (Opus 4.8/4.7/4.6, Sonnet 4.6, Haiku 4.5, Fable 5) work headlessly via --model claude-… without a settings entry. Key from ANTHROPIC_API_KEY.
• provider: "openai-compatible" → any OpenAI-compatible endpoint; requires baseUrl. Key from apiKey on the entry.

​

The TUI

• Streaming responses rendered as markdown (headers, lists, code fences, inline code, links) via a lightweight ANSI renderer.
• Thinking — reasoning streams live under ✦ Thinking… and collapses to ✦ Thought for Ns when done. Ctrl+O toggles expanded thinking.
• Tool rows — each tool call shows a formatted name, a one-line summary, live “running” state, then ✓/✗ with a short result preview.
• Edit diffs — file-modifying tools render a real diff with a stats header and red/green backgrounds, both in the approval prompt and the finished tool row.
• Tasks — the model maintains a checklist via update_todo_list, rendered as a compact Tasks panel.
• @-references — type @ in the input to fuzzy-search workspace files.
• Status bar — approval mode, busy state, model name, context token usage, and session cost.

​

Approvals & safety

• File edits — prompt shows the target; y allow once, n deny, a allow for the rest of the session.
• Commands — prompt shows the exact command line. Dangerous commands (deletes, force-pushes, system changes) can never be auto-approved.

​

Headless mode

orbcode -p "explain the build pipeline in this repo"
orbcode -p "fix the lint errors" --yolo

​

MCP servers, skills & memory

• MCP servers — OrbCode connects to external tools via the Model Context Protocol. Servers are configured in three scopes (user, project .mcp.json, local) and their tools appear alongside the native tools as mcp__<server>__<tool>. Manage them with the orbcode mcp subcommand or the /mcp slash command. See the MCP servers page.
• Skills — reusable instruction sets the model loads on demand. Place a SKILL.md under ~/.orbcode/skills/ (user) or .orbcode/skills/ (project); the catalog is injected into the system prompt and use_skill loads the full body when a task matches. See the Skills & memory page.
• AGENTS.md memory — project- and user-level instructions (build commands, code style, conventions) injected into every system prompt. OrbCode’s equivalent of Claude Code’s CLAUDE.md, using the open AGENTS.md filename. The repo-level file lives at .orb/AGENTS.md — the same folder the Orbital IDE reads, so a single source of truth covers terminal and editor. Use /init to generate or refine a starter file.
• Linked repositories (/link) — link other repos on your machine by folder path (absolute, ~/path, or relative to the project). Links are persisted in .orb/links.json (also read by the IDE), and each linked repo’s AGENTS.md is pulled in ahead of time so the model can see and act on it.

​

Architecture