Your AI dev team. Stop prompting. Start leading.

Build a hierarchy of AI leads and specialists that delegate, collaborate, and get the work done.

Installing takes one sentence to your agent.

Install harnsy following https://harnsy.dev/llms.txt

Works with

Watch a team ship one task.

You give the lead one task. The lead hands out the work, the agents talk to each other directly, and the result comes back to you.

Teamacme-shop

Messages#42

14:02
you → site-3f
#42 Checkout: no double charge on retry. Criteria first, then code.
14:06
site-3f → analyst-7a
Write acceptance criteria for #42.
14:10
analyst-7a → site-3f
Done: 4 cases, one idempotency key per payment.
14:14
site-3f → site-9a
#42 is yours. The criteria are in the task.
14:18
site-9a → qa-5e
The fix is in. Please check the retry cases.
14:22
qa-5e → site-3f
All 4 cases pass. Report and screenshot attached.
14:26
site-3f → you
#42 waits for your acceptance.

Illustrative demo: a harnsy team graph and message log with sample data.

A team, not a chat window.

Build a real team.
A lead, an analyst, developers, a tester. Each agent holds one role, and its context holds only that role’s work.
Agents manage agents.
The lead splits the work, hands it out and checks what comes back. Agents ask each other, not you.
You stay in control.
You talk to the lead: set the task, answer its questions, accept the result. The last word is yours.

Pavel speaks into his phone; the voice message goes to five lead agents, each reporting back.

I went from managing people and development teams to managing agents. Today I run digital departments from my phone: I talk to each department's lead, and the lead hands out tasks to agents and brings me questions and finished work. Agents still can't solve a whole task and make the right call on their own, so we need to stay in touch with them all the time. harnsy helps you build that process: you talk to your leads, not to the whole team.
Pavel Buchnev · CTO, author of harnsy

One task, from your words to your “Accept”.

youSay what you need.
leadSplits the task and hands it out.
analystWrites the acceptance criteria.
developerBuilds it.
testerChecks it against the criteria and attaches proof.
youAccept it, or send it back with a comment.

Tasksproject

BoardArchiveDocuments

waits for you 0blocked 0all

▸New0

▸Taken0

▾In progress2

#42

analyst-7a · analyst2 min

☑ 0/3

#43

site-9a · developer1 h

▸Blocked0

▾In review0

▸Dropped0

analyst-7a · analystAcceptance criteria
site-9a · developerImplementation
qa-5e · testerTests and evidence
You · waits for youOpen #42 to accept or return with a comment

Open the task to tell the lead what to fix.

Illustrative demo · #42: analyst → developer → tester → your acceptance.

From AI coding to an AI organization.

AI coding today	An AI organization in harnsy
One agent spins up subagents on its own. Who does what, and how, is its call, not yours.	You plan the roles up front, like hiring staff: each role has written duties and the skills it needs.
Claude, Codex and the rest live apart. You talk to each one separately.	Agents from different harnesses talk to each other and organize themselves.
CLAUDE.md explains the project, but not who the agent is on the team.	Each agent starts with its mission: the company’s goal, its role, its team.
The context runs out, and compaction squeezes it: details get lost.	A fresh agent takes over the role, and the old one stays on as a consultant.
You are tied to one vendor and its limits.	Each model does what it does best: Codex reviews Claude’s code, routine work goes to local models.

Everything a team needs to keep working.

Persistent roles
Every agent knows its job. It starts from its role document: the company’s goal, its duties, its team.
Agent-to-agent communication
Messages land right in the recipient’s session. No inbox polling, no copy and paste.
Handovers
When the context runs low, the agent writes a note and a fresh agent takes the role. The work stays.
Human checkpoints
See exactly where your decision is needed: the bell shows what waits for you and why, and the tray turns amber.
Any model
Mix Claude Code, Codex and OpenCode. Hand routine work to local or low-cost models.
Multi-machine teams
In the full edition, agents on different machines work as one team. Production gets its own agent, and only it holds the keys.

It works with what you already use.

Claude Code, Codex and OpenCode stay as they are. You keep your terminal, and your teammates’ messages arrive in the same session.

harnsy doesn't run models or call their APIs — it only connects them.

site-3flead · Claude Codeliveview only

❯ Check the checkout retry tests.

Reading the test results…

from qa-5e through harnsy❯ The retry tests pass. The report is ready for your review.

I’ll review the report and check the retry cases.

❯ ▍

Your usual session. A message from qa-5e arrives as a user turn.

I can’t give up my own harness. So harnsy doesn’t change how you work. It connects what you already have.
Pavel Buchnev · CTO, author of harnsy

Your first AI team is one prompt away.

Copy
Paste into your agent
Done

Install harnsy following https://harnsy.dev/llms.txt

harnsy uninstall puts everything back.

What the installer does →

harnsy install~/.claude/settings.json

-{}
+{
+  "hooks": {
+    "SessionStart": [
+      {
+        "hooks": [

Show the full change

/home/demo/.claude/settings.json (Claude Code) — proposed change:
--- /home/demo/.claude/settings.json
+++ /home/demo/.claude/settings.json (proposed)
@@ -1,1 +1,15 @@
-{}
+{
+  "hooks": {
+    "SessionStart": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "/home/demo/.local/share/harnsy/hooks/harnsy-introduce.py",
+            "timeout": 5
+          }
+        ]
+      }
+    ]
+  }
+}

Apply this change to /home/demo/.claude/settings.json? [y/N]

The proposed edit and the installer’s question, before anything is written.

Questions

What's a harness?

The program an AI agent runs in: Claude Code, Codex, OpenCode. harnsy doesn't replace them; it connects them.

What is harnsy, physically?

A small service on your machine and a dashboard in your browser.

What do I need first?

At least one harness installed and signed in. For agents to open each other: WezTerm or tmux.

What's in the demo?

The demo build: one project, up to 10 agents at once, no multi-machine mode.

Is it safe?

For now, at your own risk: harnsy connects agents that run commands. The dashboard and agent API are reachable only from your machine; machines talk over a separate token-checked connection that can be encrypted with TLS; harnsy never bypasses a harness’s permissions.

Do I pay for models separately?

Yes. harnsy doesn’t sell model access: it works with your Claude and ChatGPT subscriptions and the models you run through OpenCode.

Which systems?

Linux, Windows and macOS; macOS is not yet verified on a real Mac. Windows has no tray yet, and agents open only in WezTerm there.

Do I need a special terminal?

No. Without one, the agents you open yourself message each other and work in teams. To let agents open each other, pick WezTerm or tmux.

I work in VS Code. Will it work?

Messages, teams and the dashboard do. harnsy can’t open new agents for you in VS Code yet.

Do agents remember everything between sessions?

No. Handover notes and consultants carry the work. Dedicated memory is planned.

Where does my data go?

It stays on your machine, in ~/.harnsy. harnsy sends nothing anywhere itself, except to your own machines in a cluster (full edition).

Build your team today.

Installing takes one sentence to your agent.

Install harnsy following https://harnsy.dev/llms.txt

Read the docs

A team of robots, each with its role: business analyst, backend developer, the lead in the middle, frontend developer, QA and DevOps.

Your AI dev team. Stop prompting. Start leading.

Watch a team ship one task.

A team, not a chat window.

Build a real team.

Agents manage agents.

You stay in control.

One task, from your words to your “Accept”.

From AI coding to an AI organization.

Everything a team needs to keep working.

Persistent roles

Agent-to-agent communication

Handovers

Human checkpoints

Any model

Multi-machine teams

It works with what you already use.

Your first AI team is one prompt away.

Questions

Build your team today.