Hi, I'm TTal — yes, a snail. Slow to break things, fast to ship them.
I carry my whole house in one binary — no cloud, no containers, no dependencies sprawl. Drop me in your terminal and I'll route tasks to the right agent, spawn workers in parallel, and leave a trail of merged PRs behind me.
Agent ops CLI. One binary. Your terminal is the factory floor.
TTal turns your terminal into an agent assembly line. You describe what needs to happen, TTal handles the rest:
- Route tasks to the right agent — researcher investigates, designer plans, worker implements
- Spawn workers in isolated git worktrees — parallel execution, zero conflicts
- Ship PRs with automated review — 6 review agents, squash-merge from your phone
- Talk to your agents from Telegram — they talk back
task created → research → design → worker spawns → PR opens → review → merge → cleanup
All automatic. All in your terminal.
# Create a task
ttal task add --project myapp "Add JWT authentication to the API"
# Route to researcher — she investigates options, writes findings
ttal task route abc12345 --to athena
# Route to designer — he reads research, writes the implementation plan
ttal task route abc12345 --to inke
# Execute — worker spawns in its own worktree, follows the plan, opens a PR
ttal task advance abc12345
# Meanwhile, you're on your phone
# Review agents post verdicts, worker triages feedback, you merge from TelegramWe built TTal with TTal. 356 PRs merged, 29k lines of Go in 33 days. Then we pointed it at flicknote-cli — 55 PRs merged in 15 days, Rust. Same pipeline, different repo, same velocity.
TTal is three things:
┌─────────────────────────────────────────┐
│ TTal the orchestrator │
│ routes tasks, spawns │
│ workers, manages agents │
├─────────────────────────────────────────┤
│ logos the reasoning engine │
│ bash-only agent loop │
│ LLMs think in plain text │
├─────────────────────────────────────────┤
│ temenos the sacred boundary │
│ seatbelt (macOS) / bwrap │
│ (Linux) — YAGNI containers│
└─────────────────────────────────────────┘
TTal coordinates. logos thinks. temenos isolates. Workers can't touch each other or the host.
Logos is a bash-only reasoning engine — no tool schemas, no JSON ceremony. Temenos is OS-native filesystem isolation — no containers needed. Three repos, one pipeline.
brew tap tta-lab/ttal
brew install ttalOr from source:
go install github.com/tta-lab/ttal-cli@latestClone the repo and run /setup in Claude Code:
git clone https://github.com/tta-lab/ttal-cli.git && cd ttal-cli
# Open in Claude Code, then: /setupThe setup skill installs TTal, configures hooks, and walks you through Telegram integration. Five minutes to your first automated PR.
Or do it manually:
ttal doctor --fix # install hooks
ttal daemon install # start the communication hubTTal agents aren't chatbots. They're specialists with clear roles:
| Agent | Role | What they do |
|---|---|---|
| Yuki 🐱 | Orchestrator | Routes tasks, manages the pipeline |
| Athena 🦉 | Researcher | Investigates problems, writes findings |
| Inke 🐙 | Designer | Reads research, writes implementation plans |
| Workers | Coders | Spawn per-task, implement, open PRs, self-cleanup |
Each agent runs in its own tmux session. Workers get isolated git worktrees — they can't step on each other. The daemon handles all messaging: Telegram in, agent-to-agent routing, status updates out.
TTal is the execution layer of the GuionAI ecosystem:
- FlickNote captures knowledge — voice memos, links, meeting notes
- TTal agents read and write to FlickNote via CLI — plans, research findings, even agent definitions live there
No MCP. CLI-first. Agents use the same tools you do.
MIT