Manager Plane β Long-running agents with specialized roles. Researcher, designer, orchestrator. They persist across sessions, have memory, and coordinate via agent-to-agent messaging.
Worker Plane β Short-lived coders and reviewers. One per task, isolated in git worktrees. Multiple workers run in parallel. They implement, review, triage, and merge β then they're done.
TTAL doesn't replace your coding agent β it makes it a team player.
Capability
TTAL
Paperclip
Multi-agent coordination
β
β
Zero infrastructure (no database)
β
-
Bidirectional Telegram
β
-
Autonomous PR workflow
β
-
The closest competitor is Paperclip (12K stars) β multi-runtime support, goal hierarchy, React dashboard. It requires PostgreSQL and a Node.js server. TTAL is a single Go binary with no database. Paperclip models your team as an org chart; TTAL models it as two planes β coordinated through git-native workflows and direct Telegram access.
Competitors build chat assistants or company simulators. TTAL builds autonomous software teams who own the full delivery pipeline.
ttal is a CLI that orchestrates coding agents into an autonomous software team. It handles agent lifecycle, task routing, Telegram messaging, PR review pipelines, and inter-agent communication β so your agents work as a coordinated team, not isolated sessions.
Claude Code (stable) as the primary runtime. Codex CLI is supported as a worker runtime. You can mix runtimes across your team β run managers on Claude Code and workers on Codex, or any combination.
Telegram is the primary mobile interface β send messages, approve actions, check status, answer agent questions. Without it, you can still use ttal from the terminal via CLI and tmux sessions directly.
ttal adds the coordination layer: task routing, agent-to-agent messaging, persistent identity with memory, automated PR review pipelines, and a daemon that keeps everything running. Without it, you're manually managing tmux sessions and copy-pasting between agents.
The daemon is a background process that handles Telegram polling, agent lifecycle, output watching, PR status monitoring, and message routing between agents and humans. On macOS it's managed by launchd and starts automatically on boot.
The CLI works on Linux. The daemon currently uses launchd (macOS) for process management β systemd support is on the roadmap.
About 10 minutes. Clone, make install, ttal daemon install, configure your agents in config.toml, and you're running.
GitHub and Forgejo are supported natively for PR workflows (create, review, merge). GitLab support is planned.
Agent teams. A team is a group of agents that work together on a product β each team has its own agent registry, project registry, and task queue. Think of it as one team per product. Free includes 1 team with 2 agents, Pro gives 1 team with unlimited agents, Team gives unlimited teams. There are no per-seat charges for humans.
Yes, natively. Agents can message any other agent regardless of team β within-team and cross-team messaging both work out of the box via ttal send.