\n\n> New issues and PRs from new contributors are auto-closed by default. Maintainers review auto-closed issues daily. See [CONTRIBUTING.md](../../CONTRIBUTING.md).\n\n---\n\nPi is a minimal terminal coding harness. Adapt pi to your workflows, not the other way around, without having to fork and modify pi internals. Extend it with TypeScript [Extensions](#extensions), [Skills](#skills), [Rules](#rules), [Prompt Templates](#prompt-templates), and [Themes](#themes). Put your extensions, skills, rules, prompt templates, and themes in [Pi Packages](#pi-packages) and share them with others via npm or git.\n\nPi ships with powerful defaults and optional isolated subagent delegation, while leaving broader workflows such as plan mode to extensions and packages.\n\nPi runs in four modes: interactive, print or JSON, RPC for process integration, and an SDK for embedding in your own apps. See [openclaw/openclaw](https://github.com/openclaw/openclaw) for a real-world SDK integration.\n\n## Share your OSS coding agent sessions\n\nIf you use pi for open source work, please share your coding agent sessions.\n\nPublic OSS session data helps improve models, prompts, tools, and evaluations using real development workflows.\n\nFor the full explanation, see [this post on X](https://x.com/badlogicgames/status/2037811643774652911).\n\nTo publish sessions, use [`badlogic/pi-share-hf`](https://github.com/badlogic/pi-share-hf). Read its README.md for setup instructions. All you need is a Hugging Face account, the Hugging Face CLI, and `pi-share-hf`.\n\nYou can also watch [this video](https://x.com/badlogicgames/status/2041151967695634619), where I show how I publish my `pi-mono` sessions.\n\nI regularly publish my own `pi-mono` work sessions here:\n\n- [badlogicgames/pi-mono on Hugging Face](https://huggingface.co/datasets/badlogicgames/pi-mono)\n\n## Table of Contents\n\n- [Quick Start](#quick-start)\n- [Providers & Models](#providers--models)\n- [Interactive Mode](#interactive-mode)\n - [Editor](#editor)\n - [Commands](#commands)\n - [Keyboard Shortcuts](#keyboard-shortcuts)\n - [Message Queue](#message-queue)\n- [Sessions](#sessions)\n - [Branching](#branching)\n - [Compaction](#compaction)\n- [Settings](#settings)\n- [Language Servers](#language-servers)\n- [Remote Daemon](#remote-daemon)\n- [Docker Sandbox](#docker-sandbox)\n- [Customization](#customization)\n - [Prompt Templates](#prompt-templates)\n - [Skills](#skills)\n - [Rules](#rules)\n - [Extensions](#extensions)\n - [Themes](#themes)\n - [Pi Packages](#pi-packages)\n- [Programmatic Usage](#programmatic-usage)\n- [Philosophy](#philosophy)\n- [CLI Reference](#cli-reference)\n\n---\n\n## Quick Start\n\n```bash\nnpm install -g --ignore-scripts @fleetagent/pi-coding-agent\n```\n\n`--ignore-scripts` disables dependency lifecycle scripts during install. Pi does not require install scripts for normal npm installs.\n\nInstaller alternative:\n\n```bash\ncurl -fsSL https://pi.dev/install.sh | sh\n```\n\nAuthenticate with an API key:\n\n```bash\nexport ANTHROPIC_API_KEY=sk-ant-...\npi\n```\n\nOr use your existing subscription:\n\n```bash\npi\n/login # Then select provider\n```\n\nThen just talk to pi. By default, pi gives the model file and shell tools. Enable the built-in subagent tool family through `/settings` or `enableSubagents` in `settings.json`. LSP tools require both [an external language server configuration](docs/lsp.md) and `--enable-lsp-tools` (or `enableLspTools` in `settings.json`); they are off by default. The model uses enabled tools to fulfill your requests. Add capabilities via [skills](#skills), [prompt templates](#prompt-templates), [extensions](#extensions), or [pi packages](#pi-packages).\n\n**Platform notes:** [Windows](docs/windows.md) | [Termux (Android)](docs/termux.md) | [tmux](docs/tmux.md) | [Terminal setup](docs/terminal-setup.md) | [Shell aliases](docs/shell-aliases.md)\n\n---\n\n## Providers & Models\n\nFor each built-in provider, pi maintains a list of tool-capable models, updated with every release. Authenticate via subscription (`/login`) or API key, then select any model from that provider via `/model` (or Ctrl+L).\n\n**Subscriptions:**\n- Anthropic Claude Pro/Max\n- OpenAI ChatGPT Plus/Pro (Codex)\n- GitHub Copilot\n\n**API keys:**\n- Anthropic\n- OpenAI\n- Azure OpenAI\n- DeepSeek\n- Google Gemini\n- Google Vertex\n- Amazon Bedrock\n- Mistral\n- Groq\n- Cerebras\n- Cloudflare AI Gateway\n- Cloudflare Workers AI\n- xAI\n- OpenRouter\n- Vercel AI Gateway\n- ZAI\n- OpenCode Zen\n- OpenCode Go\n- Hugging Face\n- Fireworks\n- Together AI\n- Kimi For Coding\n- MiniMax\n- Xiaomi MiMo\n- Xiaomi MiMo Token Plan (China)\n- Xiaomi MiMo Token Plan (Amsterdam)\n- Xiaomi MiMo Token Plan (Singapore)\n\nSee [docs/providers.md](docs/providers.md) for detailed setup instructions.\n\n**Custom providers & models:** Add providers via `~/.pi/agent/models.json` if they speak a supported API (OpenAI, Anthropic, Google). For custom APIs or OAuth, use extensions. See [docs/models.md](docs/models.md) and [docs/custom-provider.md](docs/custom-provider.md).\n\n---\n\n## Interactive Mode\n\n
\n\nThe interface from top to bottom:\n\n- **Startup header** - Shows shortcuts (`/hotkeys` for all), loaded AGENTS.md files, prompt templates, skills, and extensions\n- **Messages** - Your messages, assistant responses, tool calls and results, notifications, errors, and extension UI\n- **Editor** - Where you type; border color indicates thinking level\n- **Footer** - Working directory, session name, total token/cache usage, cost, context usage, current model\n\nThe editor can be temporarily replaced by other UI, like built-in `/settings` or custom UI from extensions (e.g., a Q&A tool that lets the user answer model questions in a structured format). [Extensions](#extensions) can also replace the editor, add widgets above/below it, a status line, custom footer, or overlays.\n\n### Editor\n\n| Feature | How |\n|---------|-----|\n| File reference | Type `@` to fuzzy-search project files |\n| Path completion | Tab to complete paths |\n| Multi-line | Shift+Enter (or Ctrl+Enter on Windows Terminal) |\n| Clipboard | Ctrl+V to paste an image or text (Alt+V on Windows), or drag images onto terminal |\n| Bash commands | `!command` runs and sends output to LLM, `!!command` runs without sending |\n\nStandard editing keybindings for delete word, undo, etc. See [docs/keybindings.md](docs/keybindings.md).\n\n### Commands\n\nType `/` in the editor to trigger commands. [Extensions](#extensions) can register custom commands, [skills](#skills) are available as `/skill:name`, and [prompt templates](#prompt-templates) expand via `/templatename`.\n\n| Command | Description |\n|---------|-------------|\n| `/login`, `/logout` | OAuth authentication |\n| `/model` | Switch models |\n| `/scoped-models` | Enable/disable models for Ctrl+P cycling |\n| `/settings` | Thinking level, theme, message delivery, transport |\n| `/resume` | Pick from previous sessions |\n| `/new` | Start a new session |\n| `/name ` | Set session display name |\n| `/session` | Show session info (file, ID, messages, tokens, cost) |\n| `/tree` | Jump to any point in the session and continue from there |\n| `/fork` | Create a new session from a previous user message |\n| `/clone` | Duplicate the current active branch into a new session |\n| `/compact [prompt]` | Manually compact context, optional custom instructions |\n| `/context-compression [on\\|off]` | Show or change agent-initiated compression for this session (default: off) |\n| `/compress-detection [on\\|off]` | Show or change background detection for this session (default: off); enabling also enables compression |\n| `/compress-detection-model [provider/model-id\\|clear]` | Choose a detector model and enable detection for this session, or clear it |\n| `/copy` | Copy last assistant message to clipboard |\n| `/export [file]` | Export session to HTML file |\n| `/share` | Upload as private GitHub gist with shareable HTML link |\n| `/reload` | Reload keybindings, hooks, extensions, skills, rules, prompts, and context files (themes hot-reload automatically) |\n| `/hotkeys` | Show all keyboard shortcuts |\n| `/changelog` | Display version history |\n| `/quit` | Quit pi |\n\n### Keyboard Shortcuts\n\nSee `/hotkeys` for the full list. Customize via `~/.pi/agent/keybindings.json`. See [docs/keybindings.md](docs/keybindings.md).\n\n**Commonly used:**\n\n| Key | Action |\n|-----|--------|\n| Ctrl+C | Clear editor |\n| Ctrl+C twice | Quit |\n| Escape | Cancel/abort |\n| Escape twice | Open `/tree` |\n| Ctrl+L | Open model selector |\n| Ctrl+P / Shift+Ctrl+P | Cycle scoped models forward/backward |\n| Shift+Tab | Cycle thinking level |\n| Ctrl+O | Collapse/expand tool output |\n| Ctrl+T | Collapse/expand thinking blocks |\n| Ctrl+X | Copy the last assistant message |\n\n### Message Queue\n\nSubmit messages while the agent is working:\n\n- **Enter** queues a *steering* message, delivered after the current assistant turn finishes executing its tool calls\n- **Alt+Enter** queues a *follow-up* message, delivered only after the agent finishes all work\n- **Escape** aborts and restores queued messages to editor\n- **Alt+Up** retrieves queued messages back to editor\n\nOn Windows Terminal, `Alt+Enter` is fullscreen by default. Remap it in [docs/terminal-setup.md](docs/terminal-setup.md) so pi can receive the follow-up shortcut.\n\nConfigure delivery in [settings](docs/settings.md): `steeringMode` and `followUpMode` can be `\"one-at-a-time\"` (default, waits for response) or `\"all\"` (delivers all queued at once). `transport` selects provider transport preference (`\"sse\"`, `\"websocket\"`, or `\"auto\"`) for providers that support multiple transports.\n\n---\n\n## Sessions\n\nSessions are stored as JSONL files with a tree structure. Each entry has an `id` and `parentId`, enabling in-place branching without creating new files. See [docs/session-format.md](docs/session-format.md) for file format.\n\n### Management\n\nSessions auto-save to `~/.pi/agent/sessions/` organized by working directory.\n\n```bash\npi -c # Continue most recent session\npi -r # Browse and select from past sessions\npi --no-session # Ephemeral mode (don't save)\npi --session # Use specific session file or ID\npi --fork # Fork specific session file or ID into a new session\n```\n\nUse `/session` in interactive mode to see the current session ID before reusing it with `--session ` or `--fork `.\n\n### Branching\n\n**`/tree`** - Navigate the session tree in-place. Select any previous point, continue from there, and switch between branches. All history preserved in a single file.\n\n
\n\n- Search by typing, fold/unfold and jump between branches with Ctrl+←/Ctrl+→ or Alt+←/Alt+→, page with ←/→\n- Filter modes (Ctrl+O): default → no-tools → user-only → labeled-only → all\n- Press Ctrl+X to copy the selected entry\n- Press Shift+L to label entries as bookmarks and Shift+T to toggle label timestamps\n\n**`/fork`** - Create a new session file from a previous user message on the active branch. Opens a selector, copies the active path up to that point, and places the selected prompt in the editor for modification.\n\n**`/clone`** - Duplicate the current active branch into a new session file at the current position. The new session keeps the full active-path history and opens with an empty editor.\n\n**`--fork `** - Fork an existing session file or partial session UUID directly from the CLI. This copies the full source session into a new session file in the current project.\n\n### Compaction\n\nLong sessions can exhaust context windows. Compaction summarizes older messages while keeping recent ones.\n\n**Manual:** `/compact` or `/compact `\n\n**Automatic:** Enabled by default. Triggers on context overflow (recovers and retries) or when approaching the limit (proactive). Configure via `/settings` or `settings.json`.\n\nCompaction is lossy, but the full history remains in the JSONL file. The model can use `session_search` to locate finalized entries omitted from context and `session_entry_get` to retrieve one exact model-visible entry; `/tree` remains available for interactive navigation. The `compress_context` tool uses context metadata entry IDs to replace a tail—or an earlier bounded range via `endEntryId`—with a summary while retaining later messages and archiving the originals on another branch. Customize compaction behavior via [extensions](#extensions). See [docs/compaction.md](docs/compaction.md) for internals.\n\n---\n\n## Settings\n\nUse `/settings` to modify common options, or edit JSON files directly:\n\n| Location | Scope |\n|----------|-------|\n| `~/.pi/agent/settings.json` | Global (all projects) |\n| `.pi/settings.json` | Project (overrides global) |\n\nSee [docs/settings.md](docs/settings.md) for all options.\n\n### Telemetry and update checks\n\nPi has two separate startup features:\n\n- **Update check:** fetches npm metadata for `@fleetagent/pi-coding-agent` to check whether a newer Pi version exists. Disable it with `PI_SKIP_VERSION_CHECK=1`. Disabling update checks only turns off this check.\n- **Install/update telemetry:** after first install or a changelog-detected update, sends an anonymous version ping to `https://pi.dev/api/report-install`. Opt out by setting `enableInstallTelemetry` to `false` in `settings.json`, or by setting `PI_TELEMETRY=0`. This does not disable update checks; Pi may still contact the npm registry for the latest version unless update checks are disabled or offline mode is enabled.\n\nUse `--offline` or `PI_OFFLINE=1` to disable all startup network operations described here, including update checks, package update checks, and install/update telemetry.\n\n## Language Servers\n\nPi supports explicitly configured managed stdio servers, attached TCP/Unix socket/named-pipe endpoints, and host-provided connections with managed or attached semantics. It does not bundle or automatically select a language server. Untrusted project settings cannot add or positively activate transports; malformed or unreadable LSP sources fail closed without terminating Pi, and `--no-lsp` is the per-invocation recovery path. See [docs/lsp.md](docs/lsp.md) for copyable configurations, remote path mapping, project trust, workspace roots, lifecycle ownership, SDK/extension APIs, migration, and troubleshooting.\n\n---\n\n## Remote Daemon\n\n`pi --daemon` starts an integrated remote workspace daemon from the `@fleetagent/pi-coding-agent` package. The local Pi process remains the orchestrator for model calls, credentials, sessions, prompts, extensions, subagents, RPC, and UI; the daemon owns one confined workspace for canonical tools, project resources, file transfer, optional process execution, and optional colocated LSP.\n\nStart a daemon:\n\n```bash\nPI_DAEMON_TOKEN='long-random-token' pi --daemon --daemon-cwd /workspace --daemon-port 8787\n```\n\nConnect from a local Pi process:\n\n```bash\nPI_REMOTE_TOKEN='long-random-token' pi --remote ws://127.0.0.1:8787/pi/workspace\n```\n\nThe retired `@fleetagent/pi-daemon` package and `pi-daemon` binary are not supported. See [docs/daemon.md](docs/daemon.md) for topology, security, configuration, Docker, LSP/resources/extensions behavior, troubleshooting, and migration.\n\n---\n\n## Docker Sandbox\n\nInteractive `/sandbox` is the sole command for inspecting, connecting, clearing, starting, listing, and stopping workspace tool backends. It supports existing sandbox daemons and Pi-owned Docker containers. The command is user-only and appears in slash-command completion, but is hidden from model-visible command catalogs, tools, prompts, skills, and rules. Local Pi keeps model/session/UI orchestration while workspace tools/resources use the selected backend. See [docs/sandbox.md](docs/sandbox.md) for command syntax, image configuration, local base-image builds, troubleshooting, and security limits.\n\n---\n\n## Context Files\n\nPi loads `AGENTS.md` (or `CLAUDE.md`) at startup from:\n- `~/.pi/agent/AGENTS.md` (global)\n- Parent directories (walking up from cwd)\n- Current directory\n\nIf a directory contains `AGENTS.override.md`, Pi loads it instead of `AGENTS.md` or `CLAUDE.md` from that directory. Context files from other directories still layer normally, including when project instructions come from a daemon tool backend.\n\nUse for project instructions, conventions, common commands. All selected files are concatenated.\n\nDisable context file loading with `--no-context-files` (or `-nc`).\n\n### System Prompt\n\nReplace the default system prompt with `.pi/SYSTEM.md` (project) or `~/.pi/agent/SYSTEM.md` (global). Append without replacing via `APPEND_SYSTEM.md`.\n\n---\n\n## Customization\n\n### Prompt Templates\n\nReusable prompts as Markdown files. Type `/name` to expand.\n\n```markdown\n\nReview this code for bugs, security issues, and performance problems.\nFocus on: {{focus}}\n```\n\nPlace in `~/.pi/agent/prompts/`, `.pi/prompts/`, or a [pi package](#pi-packages) to share with others. See [docs/prompt-templates.md](docs/prompt-templates.md).\n\n### Skills\n\nOn-demand capability packages following the [Agent Skills standard](https://agentskills.io). Invoke via `/skill:name` or let the agent load them automatically.\n\n```markdown\n\n# My Skill\nUse this skill when the user asks about X.\n\n## Steps\n1. Do this\n2. Then that\n```\n\nPlace in `~/.pi/agent/skills/`, `~/.agents/skills/`, `.pi/skills/`, or `.agents/skills/` (from `cwd` up through parent directories) or a [pi package](#pi-packages) to share with others. See [docs/skills.md](docs/skills.md).\n\n### Rules\n\nOn-demand mandatory constraints and policies. Invoke via `/rule:name` or let the agent load them automatically when the task or files match the rule description.\n\n```markdown\n\n---\nname: typescript\ndescription: Mandatory TypeScript rules. Load before editing *.ts or *.tsx files.\n---\n\n# TypeScript Rules\n\n- No `any` unless absolutely necessary.\n- Use top-level imports only.\n```\n\nPlace in `~/.pi/agent/rules/`, `~/.agents/rules/`, `.pi/rules/`, or `.agents/rules/` (from `cwd` up through parent directories) or a [pi package](#pi-packages) to share with others. See [docs/rules.md](docs/rules.md).\n\n### Extensions\n\n
\n\nTypeScript modules that extend pi with custom tools, commands, keyboard shortcuts, event handlers, and UI components.\n\n```typescript\nexport default function (pi: ExtensionAPI) {\n pi.registerTool({ name: \"deploy\", ... });\n pi.registerCommand(\"stats\", { ... });\n pi.on(\"tool_call\", async (event, ctx) => { ... });\n}\n```\n\nThe default export can also be `async`. pi waits for async extension factories before startup continues, which is useful for one-time initialization such as fetching remote model lists before calling `pi.registerProvider()`.\n\n**What's possible:**\n- Custom tools (or replace built-in tools entirely)\n- Sub-agents and plan mode\n- Custom compaction and summarization\n- Permission gates and path protection\n- Custom editors and UI components\n- Status lines, headers, footers\n- Git checkpointing and auto-commit\n- Remote daemon and sandbox execution\n- MCP server integration\n- Make pi look like Claude Code\n- Games while waiting (yes, Doom runs)\n- ...anything you can dream up\n\nPlace in `~/.pi/agent/extensions/`, `.pi/extensions/`, or a [pi package](#pi-packages) to share with others. See [docs/extensions.md](docs/extensions.md) and [examples/extensions/](examples/extensions/).\n\n### Themes\n\nBuilt-in: `dark`, `light`. Themes hot-reload: modify the active theme file and pi immediately applies changes.\n\nPlace in `~/.pi/agent/themes/`, `.pi/themes/`, or a [pi package](#pi-packages) to share with others. See [docs/themes.md](docs/themes.md).\n\n### Pi Packages\n\nBundle and share extensions, skills, rules, prompts, and themes via npm or git. Find packages on [npmjs.com](https://www.npmjs.com/search?q=keywords%3Api-package) or [Discord](https://discord.com/channels/1456806362351669492/1457744485428629628).\n\n> **Security:** Pi packages run with full system access. Extensions execute arbitrary code, and skills can instruct the model to perform any action including running executables. Review source code before installing third-party packages.\n\n```bash\npi install npm:@foo/pi-tools\npi install npm:@foo/pi-tools@1.2.3 # pinned version\npi install git:github.com/user/repo\npi install git:github.com/user/repo@v1 # tag or commit\npi install git:git@github.com:user/repo\npi install git:git@github.com:user/repo@v1 # tag or commit\npi install https://github.com/user/repo\npi install https://github.com/user/repo@v1 # tag or commit\npi install ssh://git@github.com/user/repo\npi install ssh://git@github.com/user/repo@v1 # tag or commit\npi remove npm:@foo/pi-tools\npi uninstall npm:@foo/pi-tools # alias for remove\npi list\npi update # update pi and packages (skips pinned packages)\npi update --extensions # update packages only\npi update --self # update pi only\npi update --self --force # reinstall pi even if current\npi update npm:@foo/pi-tools # update one package\npi config # enable/disable extensions, skills, rules, prompts, themes\n```\n\nPackages install to `~/.pi/agent/git/` (git) or `~/.pi/agent/npm/` (npm). Use `-l` for project-local installs (`.pi/git/`, `.pi/npm/`). Git `@ref` values are pinned tags or commits; pinned packages are skipped by `pi update`, so use `pi install git:host/user/repo@new-ref` to move an existing package to a new ref. Git packages install dependencies with `npm install --omit=dev` by default, so runtime deps must be listed under `dependencies`; when `npmCommand` is configured, git packages use plain `install` for compatibility with wrappers. If you use a Node version manager and want package installs to reuse a stable npm context, set `npmCommand` in `settings.json`, for example `[\"mise\", \"exec\", \"node@20\", \"--\", \"npm\"]`.\n\nCreate a package by adding a `pi` key to `package.json`:\n\n```json\n{\n \"name\": \"my-pi-package\",\n \"keywords\": [\"pi-package\"],\n \"pi\": {\n \"extensions\": [\"./extensions\"],\n \"skills\": [\"./skills\"],\n \"rules\": [\"./rules\"],\n \"prompts\": [\"./prompts\"],\n \"themes\": [\"./themes\"]\n }\n}\n```\n\nWithout a `pi` manifest, pi auto-discovers from conventional directories (`extensions/`, `skills/`, `rules/`, `prompts/`, `themes/`).\n\nSee [docs/packages.md](docs/packages.md).\n\n---\n\n## Programmatic Usage\n\n### SDK\n\n```typescript\nimport { AuthStorage, InMemorySessionManager, ModelRegistry, PiAgent } from \"@fleetagent/pi-coding-agent\";\n\nconst authStorage = AuthStorage.create();\nconst modelRegistry = ModelRegistry.create(authStorage);\nconst pi = await PiAgent.create({\n sessionManager: new InMemorySessionManager(),\n authStorage,\n modelRegistry,\n});\nconst session = await pi.createAgentSession();\n\nawait session.prompt(\"What files are in the current directory?\");\nawait pi.dispose();\n```\n\nFor advanced multi-session flows, use `PiAgent` methods such as `newSession()`, `switchSession()`, `fork()`, and `importFromJsonl()`.\n\nSee [docs/sdk.md](docs/sdk.md), [docs/lsp.md](docs/lsp.md), and [examples/sdk/](examples/sdk/).\n\n### RPC Mode\n\nFor non-Node.js integrations, use RPC mode over stdin/stdout:\n\n```bash\npi --mode rpc\n```\n\nRPC mode uses strict LF-delimited JSONL framing. Clients must split records on `\\n` only. Do not use generic line readers like Node `readline`, which also split on Unicode separators inside JSON payloads.\n\nSee [docs/rpc.md](docs/rpc.md) for the protocol.\n\n---\n\n## Philosophy\n\nPi is aggressively extensible so it doesn't have to dictate your workflow. Features that other tools bake in can be built with [extensions](#extensions), [skills](#skills), or installed from third-party [pi packages](#pi-packages). This keeps the core minimal while letting you shape pi to fit how you work.\n\n**No MCP.** Build CLI tools with READMEs (see [Skills](#skills)), or build an extension that adds MCP support. [Why?](https://mariozechner.at/posts/2025-11-02-what-if-you-dont-need-mcp/)\n\n**Focused sub-agents.** Enable the built-in subagent tool family through `/settings` or `enableSubagents` in `settings.json` to delegate single, parallel, or chained tasks to fresh in-memory agent sessions with isolated conversations and shared authentication/model services. The parent can define each task's persona, response format, model, and tools dynamically or use bundled and user-defined presets. Extensions and packages can still provide alternative orchestration models.\n\n**No permission popups.** Run in a container, or build your own confirmation flow with [extensions](#extensions) inline with your environment and security requirements.\n\n**No plan mode.** Write plans to files, or build it with [extensions](#extensions), or install a package.\n\n**No built-in to-dos.** They confuse models. Use a TODO.md file, or build your own with [extensions](#extensions).\n\n**No background bash.** Use tmux. Full observability, direct interaction.\n\nRead the [blog post](https://mariozechner.at/posts/2025-11-30-pi-coding-agent/) for the full rationale.\n\n---\n\n## CLI Reference\n\n```bash\npi [options] [@files...] [messages...]\n```\n\n### Package Commands\n\n```bash\npi install [-l] # Install package, -l for project-local\npi remove [-l] # Remove package\npi uninstall [-l] # Alias for remove\npi update [source|self|pi] # Update pi and packages (skips pinned packages)\npi update --extensions # Update packages only\npi update --self # Update pi only\npi update --self --force # Reinstall pi even if current\npi update --extension # Update one package\npi list # List installed packages\npi config # Enable/disable package resources\n```\n\n### Modes\n\n| Flag | Description |\n|------|-------------|\n| (default) | Interactive mode |\n| `-p`, `--print` | Print response and exit |\n| `--mode json` | Output all events as JSON lines (see [docs/json.md](docs/json.md)) |\n| `--mode rpc` | RPC mode for process integration (see [docs/rpc.md](docs/rpc.md)) |\n| `--export [out]` | Export session to HTML |\n\nIn print mode, pi also reads piped stdin and merges it into the initial prompt:\n\n```bash\ncat README.md | pi -p \"Summarize this text\"\n```\n\n### Model Options\n\n| Option | Description |\n|--------|-------------|\n| `--provider ` | Provider (anthropic, openai, google, etc.) |\n| `--model ` | Model pattern or ID (supports `provider/id` and optional `:`) |\n| `--api-key ` | API key (overrides env vars) |\n| `--thinking ` | `off`, `minimal`, `low`, `medium`, `high`, `xhigh` |\n| `--models ` | Comma-separated patterns for Ctrl+P cycling |\n| `--list-models [search]` | List available models |\n\n### Session Options\n\n| Option | Description |\n|--------|-------------|\n| `-c`, `--continue` | Continue most recent session |\n| `-r`, `--resume` | Browse and select session |\n| `--session ` | Use specific session file or partial UUID |\n| `--fork ` | Fork specific session file or partial UUID into a new session |\n| `--session-dir ` | Custom session storage directory |\n| `--no-session` | Ephemeral mode (don't save) |\n| `--lsp-config ` | Load an external LSP layer; the file argument is startup-cwd-relative, while paths declared inside use the file's directory |\n| `--no-lsp` | Disable LSP for this invocation, including recovery from malformed lower-precedence LSP sources |\n### Tool Options\n\n| Option | Description |\n|--------|-------------|\n| `--tools `, `-t ` | Allowlist specific tool names across built-in, extension, and custom tools |\n| `--no-builtin-tools`, `-nbt` | Disable built-in tools by default but keep extension/custom tools enabled |\n| `--no-tools`, `-nt` | Disable all tools by default |\n| `--enable-lsp-tools` | Expose configured `lsp_*` tools by default |\n| `--remote ` | Run built-in workspace tools through a `pi --daemon` WebSocket backend |\n| `--remote-deferred` | Start with a deferred daemon backend and connect later with `/sandbox --attach ` |\n| `--remote-cwd ` | Stable daemon workspace cwd for deferred remote mode |\n| `--daemon` | Start the integrated remote workspace daemon instead of a normal Pi session |\nAvailable built-in tools: `read`, `bash`, `edit`, `write`, `grep`, `find`, `ls`, `websearch`, `session_search`, `session_entry_get`, `compress_context`, `subagent`, `subagent_runs`, `create_subagent`, and—when externally configured—`lsp_diagnostics`, `lsp_hover`, `lsp_definition`, `lsp_references`, `lsp_rename`, and `lsp_code_actions`. By default `read`, `bash`, `edit`, `write`, `websearch`, `session_search`, and `session_entry_get` are active. Context compression is opt-in via `/context-compression on`, `PI_CONTEXT_COMPRESSION=1`, or an explicit `--tools` allowlist. Background detection is separately opt-in via `/compress-detection on` or `PI_CONTEXT_COMPRESSION_DETECTION=1` and requires a configured detector model. Subagent and LSP tools require separate opt-ins (`enableSubagents` and `enableLspTools`/`--enable-lsp-tools` respectively), or an explicit `--tools` allowlist. Explicit allowlists may select any subset, while `--no-tools` and `--no-builtin-tools` remain authoritative.\n\n### Resource Options\n\n| Option | Description |\n|--------|-------------|\n| `-e`, `--extension ` | Load extension from path, npm, or git (repeatable) |\n| `--no-extensions` | Disable extension discovery |\n| `--skill ` | Load skill (repeatable) |\n| `--no-skills` | Disable skill discovery |\n| `--rule ` | Load rule (repeatable) |\n| `--no-rules` | Disable rule discovery |\n| `--prompt-template ` | Load prompt template (repeatable) |\n| `--no-prompt-templates` | Disable prompt template discovery |\n| `--theme ` | Load theme (repeatable) |\n| `--no-themes` | Disable theme discovery |\n| `--no-context-files`, `-nc` | Disable `AGENTS.override.md`, `AGENTS.md`, and `CLAUDE.md` context file discovery |\n\nCombine `--no-*` with explicit flags to load exactly what you need, ignoring settings.json (e.g., `--no-extensions -e ./my-ext.ts`).\n\n### Other Options\n\n| Option | Description |\n|--------|-------------|\n| `--system-prompt ` | Replace default prompt (context files and skills still appended) |\n| `--append-system-prompt ` | Append to system prompt |\n| `--tui-mode ` | TUI mode: `regular` (default) or experimental `fullscreen` |\n| `--verbose` | Force verbose startup |\n| `-h`, `--help` | Show help |\n| `-v`, `--version` | Show version |\n\n`fullscreen` keeps the transcript in an application-owned viewport above a fixed status/editor/footer dock. `regular` retains the main screen and terminal-owned scrollback. Set **TUI mode** in `/settings` for future sessions or use `--tui-mode` for one startup.\n\n### File Arguments\n\nPrefix files with `@` to include in the message:\n\n```bash\npi @prompt.md \"Answer this\"\npi -p @screenshot.png \"What's in this image?\"\npi @code.ts @test.ts \"Review these files\"\n```\n\n### Examples\n\n```bash\n# Interactive with initial prompt\npi \"List all .ts files in src/\"\n\n# Non-interactive\npi -p \"Summarize this codebase\"\n\n# Non-interactive with piped stdin\ncat README.md | pi -p \"Summarize this text\"\n\n# Different model\npi --provider openai --model gpt-4o \"Help me refactor\"\n\n# Model with provider prefix (no --provider needed)\npi --model openai/gpt-4o \"Help me refactor\"\n\n# Model with thinking level shorthand\npi --model sonnet:high \"Solve this complex problem\"\n\n# Limit model cycling\npi --models \"claude-*,gpt-4o\"\n\n# Read-only mode\npi --tools read,grep,find,ls -p \"Review the code\"\n\n# Start an integrated remote workspace daemon\nPI_DAEMON_TOKEN='long-random-token' pi --daemon --daemon-cwd /workspace --daemon-port 8787\n\n# Connect to that daemon from the orchestrating Pi\nPI_REMOTE_TOKEN='long-random-token' pi --remote ws://127.0.0.1:8787/pi/workspace\n# Start with a deferred daemon backend for later RPC or /sandbox attachment\npi --remote-deferred --remote-cwd /workspace --mode rpc\n\n# High thinking level\npi --thinking high \"Solve this complex problem\"\n```\n\n### Environment Variables\n\n| Variable | Description |\n|----------|-------------|\n| `PI_CODING_AGENT_DIR` | Override config directory (default: `~/.pi/agent`) |\n| `PI_CODING_AGENT_SESSION_DIR` | Override session storage directory (overridden by `--session-dir`) |\n| `PI_REMOTE_TOKEN` | Bearer token sent by the Pi remote-workspace client to `pi --daemon` |\n| `PI_REMOTE_SESSION_BASE_URL` | Remote session service URL (overridden by `--remote-session-base-url`) |\n| `PI_REMOTE_SESSION_TOKEN` | Bearer token for remote session service (overridden by `--remote-session-token`) |\n| `PI_REMOTE_PROJECT_ID` | Project id sent to remote session service (overridden by `--remote-project-id`) |\n| `PI_PACKAGE_DIR` | Override package directory (useful for Nix/Guix where store paths tokenize poorly) |\n| `PI_OFFLINE` | Disable startup network operations, including update checks, package update checks, and install/update telemetry |\n| `PI_SKIP_VERSION_CHECK` | Skip the Pi version update check at startup. This prevents the npm metadata request |\n| `PI_TELEMETRY` | Override install/update telemetry. Use `1`/`true`/`yes` to enable or `0`/`false`/`no` to disable. This does not disable update checks |\n| `PI_CONTEXT_COMPRESSION` | Set to `1` to enable `compress_context` by default for all agents. Detection remains off |\n| `PI_CONTEXT_COMPRESSION_DETECTION` | Set to `1` to enable background detection and `compress_context` for all agents. Requires a configured `compressionDetectionModel` |\n| `PI_CACHE_RETENTION` | Set to `long` for extended prompt cache (Anthropic: 1h, OpenAI: 24h) |\n| `VISUAL`, `EDITOR` | External editor for Ctrl+G |\n| `PI_DAEMON_HOST`, `PI_DAEMON_PORT`, `PI_DAEMON_CWD`, `PI_DAEMON_TOKEN` | Integrated daemon bind address, port, workspace root, and bearer token |\n| `PI_DAEMON_TEMP_ROOT` | Optional additional confined daemon temporary root; the default sandbox image uses `/tmp` |\n---\n\n## Contributing & Development\n\nSee [CONTRIBUTING.md](../../CONTRIBUTING.md) for guidelines and [docs/development.md](docs/development.md) for setup, forking, and debugging.\n\n---\n\n## License\n\nMIT\n\n## See Also\n\n- [@fleetagent/pi-ai](https://www.npmjs.com/package/@fleetagent/pi-ai): Core LLM toolkit\n- [@fleetagent/pi-agent-core](https://www.npmjs.com/package/@fleetagent/pi-agent-core): Agent framework\n- [@fleetagent/pi-tui](https://www.npmjs.com/package/@fleetagent/pi-tui): Terminal UI components\n","readmeFilename":"README.md"}
![\"pi](\"https://pi.dev/logo-auto.svg\"){kind=logo}
\n \n![\"Discord\"](\"https://img.shields.io/badge/discord-community-5865F2?style=flat-square&logo=discord&logoColor=white\"){kind=icon}
![\"npm\"](\"https://img.shields.io/npm/v/@fleetagent/pi-coding-agent?style=flat-square\")
![\"Exy](\"docs/images/exy.png\")
![\"Interactive](\"docs/images/interactive-mode.png\")
![\"Tree](\"docs/images/tree-view.png\")
![\"Doom](\"docs/images/doom-extension.png\")