Code is cheap, show me your agentic system.
devos.ing is the agentic workflow OS. Manage every task on a project board, watch your agents run the loop, and stay in sync from Telegram - wherever you are.
$ curl -fsSL https://devos.ing/cli | sh★ RUNS LOCALLY ★ YOUR MACHINE ★ YOUR KEYS ★
▓ GET STARTED ▓
Run the command.
Onboard in 30 seconds.
One command bootstraps your workspace, links your repo, and spins up your first agent crew. No installs, no config files.
▓ HOW IT WORKS ▓
One flow. Five moving parts.
From a one-line idea to a merged PR - every step is wired up. You drop the work in, devos picks it up.
- 01Drop the idea
Devs and PMs file tasks or ideas directly on devos.ing - one line is enough.
- 02Daemon polls
The devos daemon polls the shared task queue from the server DB on a tight loop.
- 03Pipeline runs
Daemon dispatches agents through explore -> plan -> implement -> test -> review.
- 04Talk in Telegram
Check status, approve, pause, or redirect - every notification is actionable.
- 05Bot ships PR
The GitHub bot reviews the diff and auto-merges when checks are green.
▓ MEET THE CREW ▓
Five bots. One pipeline.
Each agent has one job. Pick one to read its dossier.
Reads the room before anyone codes.
Crawls the repo, docs, and prior tickets to build a citation-grade context bundle. Refuses to hallucinate.
Project board
Manage agents
like a team.
Every task is a card. Every card has a status, an owner, and a trail. Drag work between humans and agents - the board is the source of truth.
- Group by goal, owner, or agent
- One-click handoff between agent and human
- Full run history with replayable trajectories
- Real-time sync to Linear, Jira, GitHub
Telegram inbox
Your agents
in your pocket.
Check status, approve PRs, and unblock agents from anywhere. devos.ing ships with a first-class Telegram bot - every notification is actionable.
/statusLive state of any task or agent/inboxEverything waiting on your approval/pauseStop an agent mid-loop, with context preserved/resumeRestart from any checkpoint
The platform
Everything your agents need. Nothing they don't.
The agentic loop
Explore -> plan -> implement -> test -> loop. A disciplined cycle every agent runs, end to end.
Project board
A shared board where humans and agents pick up tasks. Status, owner, and trail on every card.
Telegram inbox
Status, approvals, and pause/resume in chat. Every notification is actionable from your phone.
Agent runtime
Isolated, stateful runs with sub-second cold starts. Scale from one to thousands in parallel.
Branching & replay
Every run is a checkpointed graph. Branch, replay, and diff trajectories like git for behavior.
Policy & guardrails
Per-agent permissions, spending caps, and human-in-the-loop approvals. SOC 2 + HIPAA ready.
▓ SKILL TREE ▓
Skills the crew can equip.
Every skill is a typed tool with a sandbox and an audit log. Agents pick the ones they need per run - no agent gets root-level access by default.
Give your team
a thousand teammates.
Runs on your machine. Your code, your keys, your control.