Trenchsi 🪙🤖

Trenchsi is the trench rig for memecoin operators, crypto degens, builders, and power users who want an AI agent that moves fast and stays useful.

It gives you terminal chat, model selection, workspace memory, skills, scheduled jobs, and gateway-driven channel integrations so you can keep the alpha flowing without living in five different tools.

This repo contains the main project landing page plus the agent core in trenchsi/.

Who Uses It

• Solo traders who want an AI sidekick in the terminal 🤖
• Memecoin trenchers who need fast operator workflows 🪙
• Crypto builders wiring agent automation into their stack ⚙️
• Community leads running Telegram, Discord, Slack, and other channels 📣
• Tinkerers who want reusable skills, memory, and scheduled jobs 🧠

What The Agent Can Do

• Chat directly in the terminal with a configurable default model 💬
• Launch in three modes: terminal agent, web console, and TUI dashboard 🖥️
• Connect to external channels such as Telegram, Discord, Slack, Matrix, LINE, QQ, Weixin, DingTalk, Feishu, and more 🌐
• Use multiple model providers through a model-centric configuration 🔌
• Manage reusable skills for specialized workflows 🛠️
• Keep workspace context with files such as AGENT.md, USER.md, SOUL.md, and MEMORY.md 📚
• Schedule recurring work with built-in cron commands ⏰
• Run as a gateway for always-on or multi-channel agent operation 🚪

Repository Layout

• README.md: project landing page and quick start
• trenchsi/: agent source code, docs, workspace scaffold, and packaging
• trenchsi/docs/: provider, channel, and architecture docs
• trenchsi/workspace/: default workspace identity, memory, and tool instructions

Quick Start 🚀

Shortest path:

1. Build the CLI
2. Run onboarding
3. Set a default model
4. Start the agent

1. Build the CLI

From the repository root:

cd trenchsi
go build -o trenchsi ./cmd/trenchlaw

2. Run onboarding

./trenchsi onboard

This creates the local config and workspace files. install is an alias:

./trenchsi install

3. Set a default model

You can do this during onboarding or later with the CLI:

./trenchsi auth login
./trenchsi auth status
./trenchsi model
./trenchsi model gpt-5.4

If you prefer editing config directly, add a model entry such as:

{
  "model_name": "gpt-5.4",
  "model": "openai/gpt-5.4",
  "api_key": "sk-..."
}

Provider setup details are in trenchsi/docs/providers.md.

4. Start the agent

For normal use:

./trenchsi

If onboarding is complete, the interactive launcher lets you choose:

• Terminal Agent
• Web Console
• TUI Dashboard

If you want the terminal agent directly:

./trenchsi agent
./trenchsi agent -m "Summarize the repository and suggest next steps"

Useful checks:

./trenchsi status
./trenchsi version

Most Used Commands ⚡

Use these first before exploring the rest of the CLI:

./trenchsi onboard      # first-time setup
./trenchsi              # launcher: agent, web, or tui
./trenchsi agent        # terminal agent directly
./trenchsi auth login   # connect a provider
./trenchsi model        # list or set the default model
./trenchsi status       # confirm config and runtime state
./trenchsi gateway      # run integrations and long-lived channels
./trenchsi skills list  # inspect installed skills
./trenchsi cron list    # inspect scheduled jobs

Interfaces 🎛️

trenchsi is the main entry point.

• trenchsi: opens the startup selector in an interactive terminal
• trenchsi agent: skips the selector and starts terminal chat
• trenchsi gateway: runs the always-on gateway for channels and automations

If setup is incomplete, trenchsi sends you to onboarding first. If no default model is configured, trenchsi agent tells you to finish setup or choose a model.

Channel Integrations 📡

Trenchsi can operate across external messaging platforms through the gateway.

Documented channel:

• Telegram

Start the gateway with:

./trenchsi gateway

Channel setup docs:

• trenchsi/docs/channels/telegram/README.md
• trenchsi/docs/channels/discord/README.md
• trenchsi/docs/channels/slack/README.md

Skills, Memory, And Workspace Customization 🧬

The workspace is part of the agent design. Trenchsi ships with editable workspace files that define identity, memory, tools, and user-specific context.

Important files:

• trenchsi/workspace/AGENT.md
• trenchsi/workspace/USER.md
• trenchsi/workspace/SOUL.md
• trenchsi/workspace/TOOLS.md
• trenchsi/workspace/memory/MEMORY.md

Skill management commands:

./trenchsi skills list
./trenchsi skills search git
./trenchsi skills install <github-repo>
./trenchsi skills remove <skill-name>

Scheduling And Automation ⏳

Trenchsi includes built-in cron support for recurring work.

Examples:

./trenchsi cron list
./trenchsi cron add
./trenchsi cron enable <job-id>
./trenchsi cron disable <job-id>
./trenchsi cron remove <job-id>

Documentation 📖

• Provider setup: trenchsi/docs/providers.md
• Hooks: trenchsi/docs/hooks/README.md
• Agent refactor notes: trenchsi/docs/agent-refactor/README.md
• Chat app docs: trenchsi/docs/chat-apps.md

Notes 📝

• The package metadata inside trenchsi/ still references trenchlaw in several places. The repo name and top-level presentation use Trenchsi.
• .DS_Store files are currently modified in the repository and were not changed by this README update.