snscoder is a personal, single-user AI coding agent CLI. Bring your own API key, talk to the agent, and it configures itself — MCP servers, memory backends, model switching, all through conversation. Forked from Pi Agent / oh-my-pi and stripped to a focused, local-first, terminal-first agent.

Table of Contents

• Features
• Tools
• Slash Commands
• Memory Backends
• Installation
• Quick Start
• Configuration
• CLI Reference
• MCP Integration
• Telegram
• Architecture
• Development
• Contributing
• Security
• License
• Credits

Features

| Feature | Source | |---------|--------| | 30 built-in tools | src/tools/builtin-names.ts | | 157 built-in slash commands | src/slash-commands/builtin-registry.ts | | Multi-provider LLM | OpenAI, Anthropic, Ollama, custom endpoints via @oh-my-pi/pi-ai | | 7 memory backends | mnemopi (default), hindsight, mnemosyne, mem0, lcm, local, off — src/memory-backend/resolve.ts | | MCP integration | 22 source files in src/mcp/ | | Plan mode | src/plan-mode/ — agent plans before executing | | Goal mode | Autonomous objective with token budget and lifecycle | | Subagent delegation | src/task/ — spawn child agents for parallel work | | Advisor | Second model reviews each turn and injects notes | | Collaborative sessions | src/collab/ — host/join sessions via link or QR code | | Skills & plugins | src/extensibility/ — markdown skills + plugin marketplace | | Auto-learning | src/autolearn/ — agent learns from interactions | | Browser automation | Puppeteer-based headless browser — src/tools/ browser tool | | LSP integration | Language Server Protocol — src/lsp/ | | Eval backends | Python, Ruby, Julia, JavaScript — src/eval/ | | SSH remote execution | src/ssh/ | | Text-to-speech | 10 files in src/tts/ | | TinyLLM local inference | 7 files in src/tiny/ | | Telegram bot | src/adapters/telegram/ — auto-boot with SNS_TELEGRAM_BOT_TOKEN | | Debug Adapter Protocol | src/dap/ | | Context compaction | Automatic context window management with multiple strategies | | IRC | IRC tool for chat protocol interaction | | Todo system | Markdown-based todo list with reminders |

Tools (30 Built-in)

All tool names come from src/tools/builtin-names.ts.

File & Code Operations

| Tool | Description | |------|-------------| | read | Read file contents with line ranges | | write | Write / create / append to files | | edit | Find-and-replace edits in files | | find | Find files by name or pattern | | search | Search file contents (regex) |

Shell & Execution

| Tool | Description | |------|-------------| | bash | Execute shell commands | | eval | Evaluate code snippets (Python, JS, Ruby, Julia) | | ssh | Execute commands on remote hosts via SSH |

Code Intelligence

| Tool | Description | |------|-------------| | ast_grep | AST-based code search | | ast_edit | AST-aware code editing | | lsp | Language Server Protocol operations | | debug | Debugging assistance |

Web & Browser

Web & Browser

| Tool | Description | |------|-------------| | web_search | Web search via configured provider | | search_tool_bm25 | BM25-based search tool | | browser | Headless browser automation (Puppeteer) | | inspect_image | Image analysis |

Memory & Learning

| Tool | Description | |------|-------------| | memory_edit | Edit memory entries | | retain | Store information in memory | | recall | Retrieve information from memory | | reflect | Self-reflection on past actions | | learn | Learn from interactions | | manage_skill | Manage skill files |

Sessions & Tasks

| Tool | Description | |------|-------------| | task | Spawn subagent for delegated work | | job | Manage background jobs | | checkpoint | Save session checkpoint | | rewind | Rewind to a previous checkpoint |

Communication & Other

| Tool | Description | |------|-------------| | irc | IRC chat protocol | | todo | Markdown todo list management | | github | GitHub operations (repos, issues, PRs) | | ask | Ask clarifying questions |

Slash Commands

61 built-in commands from src/slash-commands/builtin-registry.ts. Most useful:

Session & Navigation

| Command | Description | |---------|-------------| | /compact | Compact conversation context | | /context | Show context-window breakdown | | /copy | Copy last response to clipboard | | /export | Export session to file | | /dump | Dump raw session transcript | | /share | Share session via collab link | | /reset | Reset session | | /help | Show help |

Model & Provider

| Command | Description | |---------|-------------| | /model | Switch model for this session | | /switch | Quick model switch (same as alt+p) | | /fast [on\|off\|status] | Toggle priority service tier | | /settings | Open settings menu | | /setup | Open provider setup wizard |

Goals & Planning

| Command | Description | |---------|-------------| | /plan | Toggle plan mode | | /plan-review | Review the active plan | | /goal <set\|show\|pause\|resume\|drop\|budget> | Goal lifecycle management | | /guided-goal | Guided goal walkthrough | | /loop [count\|duration] | Toggle loop mode | | /advisor [on\|off\|status\|dump] | Advisor subagent control |

Output & Collaboration

| Command | Description | |---------|-------------| | /collab [view\|status\|stop] | Host a live collab session | | /join <link> | Join a collab session | | /leave | Leave collab session | | /todo <edit\|copy\|export\|import\|append\|start\|done\|drop> | Todo list management | | /browser [headless\|visible] | Switch browser mode |

System

| Command | Description | |---------|-------------| | /mcp [reload] | MCP server status / reload | | /ssh | Run command on remote host | | /plugins | Show installed plugins | | /marketplace | Open marketplace manager | | /stats [--port] | Launch stats dashboard | | /usage [show\|reset] | Token usage / rate-limit reset | | /version | Show version | | /changelog [full] | Show changelog entries | | /debug | Toggle debug logging | | /theme | Switch UI theme | | /unpin | Unpin pinned context |

Memory Backends

Seven backends, configured via memory.backend in settings. Source: src/memory-backend/resolve.ts.

| Backend | Description | Source | |---------|-------------|--------| | mnemopi (default) | SQLite + vector embeddings + graph. Local, zero setup. | src/mnemopi/ (7 files) | | hindsight | Remote memory backend | src/hindsight/ (5 files) | | mnemosyne | Three-tier memory: episodic → semantic → procedural. Auto-consolidation. | src/memory-backend/mnemosyne-backend.ts | | mem0 | Semantic facts with auto-extraction. Relevance + recency scoring. | src/memory-backend/mem0-backend.ts | | lcm | Latent Context Memory. Delta encoding + multi-resolution. | src/memory-backend/lcm-backend.ts | | local | Rollout summary pipeline | src/memory-backend/local-backend.ts | | off | Memory disabled | src/memory-backend/off-backend.ts |

Cron Scheduler

Persistent background scheduler, SQLite-backed. Source: src/cron/.

/cron list              # List all jobs
/cron add <name> <expr> <action> [args]   # Add: action = prompt|shell|skill
/cron run <name>        # Run immediately
/cron remove <name>     # Delete
/cron status            # Show scheduler status
/cron enable <name>     # Enable
/cron disable <name>    # Disable

Example: /cron add backup "0 2 * * *" shell "git push" runs git push daily at 2 AM.

mnemopi Settings

| Key | Default | Description | |-----|---------|-------------| | mnemopi.dbPath | — | SQLite database path | | mnemopi.bank | — | Memory bank | | mnemopi.scoping | — | Scoping mode | | mnemopi.embeddingVariant | — | Embedding model variant | | mnemopi.autoRecall | — | Auto-recall on context load | | mnemopi.autoRetain | — | Auto-retain after conversations | | mnemopi.retainEveryNTurns | 4 | Turns between auto-retains | | mnemopi.recallLimit | 8 | Max recall results | | mnemopi.recallContextTurns | 3 | Context turns for recall | | mnemopi.recallMaxQueryChars | 4000 | Max query characters | | mnemopi.injectionTokenLimit | 5000 | Token budget for injected memories | | mnemopi.debug | false | Debug logging |

Installation

Option 1: One-liner (Linux / macOS / WSL)

curl -fsSL https://raw.githubusercontent.com/Reihantt6/sns-myagent/main/install.sh | bash

Multi-arch: detects Linux x64, Linux ARM64 (Termux/Android), macOS.

Option 2: npm (all platforms)

npm install -g @sns-myagent/cli

Works on Linux, macOS, Windows, WSL. Requires Node.js 18+. Prebuilt binary downloaded on install.

Option 3: Build from source

git clone https://github.com/Reihantt6/sns-myagent.git
cd sns-myagent
bun install        # requires Bun >= 1.3.14
bun run build      # produces bin/snscoder-linux-x64

Option 4: Windows PowerShell

irm raw.githubusercontent.com/Reihantt6/sns-myagent/main/install.ps1 | iex

Uses npm under the hood. WSL also supported via Option 1 or 2.

Run from source (development)

bun run src/cli/entry.ts

Platform Support

| Platform | Method | |----------|--------| | Linux x64 | Binary, npm, curl, source | | Linux ARM64 (Termux/Android) | Binary, npm, curl, source | | macOS x64 | Binary (CI), npm, source | | macOS ARM64 (M1/M2/M3) | Binary (CI), npm, source | | Windows x64 | Binary (CI), npm, PowerShell installer, WSL | | Android (Termux) | Binary, curl installer |

Prebuilt binaries: GitHub Releases. Auto-built for all platforms on every v* tag via .github/workflows/build-release.yml.

Quick Start

1. Set an API key

export OPENAI_API_KEY="sk-..."
# or
export ANTHROPIC_API_KEY="sk-ant-..."

2. Run

snscoder

Or from source:

bun run src/cli/entry.ts

3. Use it

> what files are in the current directory?
> search the web for "bun runtime benchmarks"
> create a TypeScript module that parses CSV files
> refactor src/utils.ts to use async/await

4. Configure through conversation

> add MCP filesystem for /home/user/projects
> switch to anthropic with claude-sonnet
> load coding skill

Configuration

Config lives at .sns-myagent/config.json (JSON, not YAML). Source: src/config/loader.ts.

Key Config Categories

Settings use dot-separated paths. Full schema in src/config/settings-schema.ts.

{
  // Model selection
  "model.*": "...",

  // Feature toggles
  "advisor.enabled": true,
  "bash.enabled": true,
  "browser.enabled": true,
  "browser.headless": true,
  "checkpoint.enabled": true,
  "compaction.enabled": true,
  "debug.enabled": false,
  "github.enabled": true,
  "goal.enabled": true,
  "todo.enabled": true,

  // Eval backends
  "eval.py": true,
  "eval.js": true,
  "eval.rb": true,
  "eval.jl": true,

  // Memory
  "memory.backend": "mnemopi",  // "mnemopi" | "hindsight" | "mnemosyne" | "mem0" | "lcm" | "local" | "off"
  "hindsight.scoping": "...",

  // MCP
  "mcp.*": {},

  // Session
  "session.*": {},

  // Theme
  "theme.dark": "titanium",

  // Compaction
  "compaction.enabled": true,
  "compaction.strategy": "...",
  "compaction.thresholdPercent": 80,
  "compaction.idleEnabled": true,
  "compaction.idleTimeoutSeconds": 300,

  // Todo
  "todo.enabled": true,
  "todo.reminders": true,
  "todo.reminders.max": 5,

  // Mnemopi (when memory.backend = "mnemopi")
  "mnemopi.dbPath": "...",
  "mnemopi.autoRecall": true,
  "mnemopi.autoRetain": true,
  "mnemopi.retainEveryNTurns": 4,
  "mnemopi.recallLimit": 8
}

Access in TUI via /settings or through conversation.

CLI Reference

snscoder                          # Interactive TUI mode
snscoder "prompt"                 # Single-prompt mode
snscoder telegram                 # Start Telegram adapter
snscoder --help                   # Show help
snscoder --version                # Show version

MCP Integration

Connect Model Context Protocol servers for additional tools. 22 source files in src/mcp/.

Through Conversation

> add MCP filesystem for /home/user/projects
> add MCP github

Through Config

Configure via /mcp or the settings panel. Servers are defined in the MCP config section.

Popular MCP Servers

| Server | Package | |--------|---------| | Filesystem | @modelcontextprotocol/server-filesystem | | PostgreSQL | @modelcontextprotocol/server-postgres | | GitHub | @modelcontextprotocol/server-github |

Telegram

Telegram bot adapter in src/adapters/telegram/ (4 files). Built on grammY.

Setup

1. Create a bot via @BotFather on Telegram
2. Set the token:

export SNS_TELEGRAM_BOT_TOKEN="your-bot-token-here"

3. Run snscoder — the bot auto-starts polling:

snscoder

The adapter starts automatically when SNS_TELEGRAM_BOT_TOKEN is set. Disable auto-start with:

export SNS_TELEGRAM_AUTOSTART=0

Or start the Telegram adapter explicitly:

snscoder telegram

Architecture

sns-myagent/
├── src/
│   ├── cli/                    # CLI entry point + commands
│   ├── config/                 # Settings schema, loader, defaults
│   ├── tools/                  # 30 built-in tools
│   ├── modes/                  # Interactive mode, TUI, controllers
│   ├── session/                # Session management
│   ├── mcp/                    # MCP integration (22 files)
│   ├── prompts/                # System prompts
│   ├── memory-backend/         # Memory backend resolver
│   ├── mnemopi/                # mnemopi backend (SQLite + vector + graph)
│   ├── hindsight/              # hindsight backend (remote memory)
│   ├── extensibility/          # Skills, plugins, marketplace
│   ├── capability/             # Skill system
│   ├── task/                   # Subagent delegation
│   ├── plan-mode/              # Plan mode
│   ├── web/                    # Web search, scrapers
│   ├── ssh/                    # SSH tool
│   ├── lsp/                    # LSP tool
│   ├── eval/                   # Eval backends (Python, Ruby, Julia, JS)
│   ├── collab/                 # Collaborative sessions
│   ├── autolearn/              # Auto-learning system
│   ├── adapters/
│   │   └── telegram/           # Telegram bot adapter
│   ├── tts/                    # Text-to-speech (10 files)
│   ├── tiny/                   # TinyLLM local inference (7 files)
│   ├── dap/                    # Debug Adapter Protocol
│   ├── stt/                    # Speech-to-text
│   ├── thinking/               # Thinking mode support
│   ├── ui/                     # Terminal UI components
│   └── utils/                  # Utilities
├── docs/                       # Documentation
├── bin/                        # Prebuilt binaries
├── scripts/                    # Build + dev scripts
├── install.sh                  # Multi-arch installer
├── install.ps1                 # Windows installer
└── package.json

Development

bun install              # Install dependencies
bun run build            # Build binary to bin/
bun run src/cli/entry.ts # Run from source
bun test                 # Run tests
bun run check            # Biome check + type check
bun run lint             # Biome lint
bun run fix              # Auto-fix lint + format

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/my-change)
3. Commit with clear messages
4. Push and open a Pull Request

Commit Convention

add:      new feature or file
fix:      bug fix
refactor: code restructuring
docs:     documentation changes
test:     test additions or changes
chore:    maintenance tasks

Security

See SECURITY.md for vulnerability reporting and security policies.

License

MIT — see LICENSE.

Implementation Status (v0.3.7)

What is actually wired and working in the source tree, verified 2026-06-30:

Tools (30 / 30) — src/tools/builtin-names.ts

All 30 tools are real implementations, not stubs.

Slash Commands (157) — src/slash-commands/builtin-registry.ts

157 commands registered. All callable via /<name> in interactive mode.

Memory Backends (7 / 7) — src/memory-backend/resolve.ts

| Backend | Status | Requires config | |---------|--------|-----------------| | mnemopi (default) | ✅ Works out of box | None — local SQLite + vector + graph | | local | ✅ Works out of box | None — rollout summary pipeline | | off | ✅ Works out of box | None — no-op | | mnemosyne | ✅ Wired | mnemosyne Python daemon running locally | | mem0 | ✅ Wired | MEM0_API_KEY (or self-hosted endpoint) | | lcm | ✅ Wired | LCM_HOST (default 127.0.0.1:8500) | | hindsight | ✅ Wired | HINDSIGHT_API_KEY or remote endpoint |

Set via memory.backend in ~/.sns-myagent/config.yaml. Default is mnemopi.

Telegram Adapter — src/adapters/telegram/

5 files (bot, handler, format, index, bridge). 10 slash commands wired: /start /help /chat /reset /status /memory /cron /model /code /review. File upload/download supported. Auto-boot when SNS_TELEGRAM_BOT_TOKEN is set.

Multi-Agent Orchestration — src/agents/

7 files. agents.yaml config, 3 ensemble strategies (consensus / critic / best-of-N), resilience (retry + circuit-breaker + timeout), 21 tests pass. CLI orchestrate <prompt> is wired but executor is stubbed (returns clear "not wired yet" message).

Terminal UI — src/tui/ + src/ui/

8 files. MY snsagent splash, ● prefix chat blocks, flat status bar, command palette, error display, memory toast. No gradient, no rounded boxes — flat list style. v0.3.6+ rebrand.

CI / CD

GitHub Actions runs 7-stage pipeline: typecheck, lint, build, test, diagnose, smoke, all-checks. 5 platform binaries (linux x64/arm64, macos x64/arm64, windows x64) auto-built on tag push. Docker multi-arch image auto-built and pushed to ghcr.io/reihantt6/sns-myagent.

Published

• npm: npm install -g @sns-myagent/cli → v0.3.7
• GitHub releases: 5 binaries per version
• Docker: ghcr.io/reihantt6/sns-myagent:v0.3.7

Credits

Forked from Pi Agent / oh-my-pi by can1357. Stripped to a focused, single-user terminal agent with conversational configuration.