Many agents, one factory
Coordinate several agents/sessions under one Chair per project, the first seat that opens. Message bus, shared locks, your declared projects and shared paths.
Orchemax
Govern multi-agent AI teams without wrecking the repo or the budget
Parallel agents need a chair, rules, and teamwork — not another chat tab. Local factory + CTO overlook; your CLIs stay native under the seat.
“As a founder or CTO, your job is to ship apps and invoice: not build the infrastructure to babysit AIs. For about a week of your cheapest engineer’s salary, get a local assembly line that is audited and 100% on your hardware.” — Buy vs build (product pitch)
Multi-agent teamwork with governance
Daily reality: several agents, several sessions, one codebase. Orchemax exists so that chaos has a chair, locks, shared reuse, and a manager overlook, not another autocomplete.
Teamwork without stomping
Agents hand off on disk, search shared before inventing, and take locks before editing common code: so parallel sessions don’t rewrite the same helpers three times.
Stop the duplicate mess
Shared reuse rules keep architecture intact while the floor runs hot. Code stays on your machines; SaaS never sees source.
How you work
Day-to-day flow only: what you do on your machine. No secret recipe; orch keeps the wiring under the seat.
1. Install & bind
Put orch in its own folder. Run orch setup and pick a separate workspace for your code. The client install is not the workshop home.
2. Open a seat
Go into a registered project folder and run orch opencode, orch claude, orch commandcode, or your agent. The first interactive seat is the Chair. A second Chair on the same project is blocked.
3. Chair vs workers
Why one Chair: if every agent acts like a director, they race the same files, invent duplicate helpers, and the work knots. Workers are for secondary tasks: directed by that Chair: not a second orchestra on the same project. Arrival order picks the Chair; you don’t appoint it in chat. Need more hands? Spawn workers; don’t open another Chair.
4. Shared code
Before editing common paths, take a lock. Edit. Release when done so teammates see the refresh. Don’t edit shared without a lock.
5. Talk on the bus
Seats leave short notes and handoffs for each other (inbox / notices). Use orch messaging: not a parallel vendor chat between agents.
6. Optional cloud
Link an Orchemax account when you want caps, spend visibility, and the CTO overlook. Source stays local; the cloud sees metadata, not your repo.
Orchemax vs other agent orchestrators
Peers are orchestrators (Orca, Traycer, OpenClaw, Hermes, Prime Agent…): not Claude or Cursor. This is a capability map: where others are ahead, we close the gap. No trophies: facts only.
Swipe sideways to read every column.
| Capability | Orca | Traycer | OpenClaw / Hermes | Prime Agent | Orchemax |
|---|---|---|---|---|---|
| Job | ADE / parallel agent IDE | Host + tasks around BYO agents | Personal multi-channel agent gateway | Self-improving coding / research harness | Governed workshop + CTO overlook |
| Keep native agent UI | Panes / terminals in ADE | Chat / Terminal Host | Own runtime face | Own TUI / daemon | orch sits under BYO CLIs |
| One director per project | Coordinator / Run model (ADE) | Task-centric Host | Usually one personal brain | Agent can spawn peers; multi-director risk | One Chair; workers only for secondary tasks |
| Stop agents stomping shared code | Worktree isolation (thin workshop law) | Artifacts / reviews: not shared locks | Not a repo factory | Not workshop locks / symbol law | Shared locks + reuse rules |
| Agent↔agent bus | Orchestration messages / gates | A2A with capability gates | Channels to humans / tools | Role-addressed messaging + roster | Disk bus + parent/child/sibling roles |
| Desktop ADE / mobile companion | Desktop ADE + mobile companion | Host / Sync app | Channel apps, not ADE | CLI / daemon focus | Console + browser overlook first |
| TG / WA command + approve | Mobile steer (product path) | Limited vs channel bots | Mature personal Telegram / WhatsApp bots | Not the core wedge | Team plan · Telegram HITL (shipping) |
| Self-improving long harness | Not the headline | Not the headline | Session continuity | RLM + Continual Harness /refine | Memory/brief + orch refine MVP (no Python RLM) |
| Chair in BYO CLI spawns workers | Coordinator / Run inside the ADE | Host + tasks (own face) | Not a multi-CLI repo factory | Can spawn peers (multi-director risk) | MCP dispatch_worker / task_* under any native CLI Chair (orch opencode|claude|commandcode|…); one Chair; ledger+policy |
| Org overlook + visible plans | Commercial opaque on pitch | BYOA + paid tiers | Personal / self-host, not org SaaS | OSS harness, not FREE→TEAM overlook | FREE→TEAM + metadata overlook |
| Source leaves your machine? | Local ADE (your hardware / SSH) | Local Host; cloud syncs task data | Self-host / your gateway | Local agent execution | Code local; SaaS metadata only |
Living matrix: updated 2026-09-13 (Chair MCP dispatch_worker / task_*). We rewrite cells when a gap closes. Complementary tools can sit beside Orchemax; the wedge is workshop law + overlook, not “another ADE.”
Workspace ≠ project
Illustration only: not a forced template. You declare paths; Orchemax binds
.orch/ + orch.yaml and meters registered projects, not folder scans.
acme-workshop/ ← WORKSPACE (not a billable project) ├── orch.yaml ├── .orch/ ← Orchemax state only ├── apps/ │ ├── api/ ← registered project │ ├── web/ │ └── worker/ ├── shared/ │ ├── go/ ← shared.go (+ orch:export) │ ├── css/ ← shared.css tip │ └── js/ ← shared.js tip ├── tools/ └── docs/ Order: 1 workspace → 2 register projects → 3 shared by language (+ assets)
Ideal order to teach: workspace root → register projects → shared by language (and cross-cutting assets when every app assumes them).
Controls for people who pay the bill
Solo builders get a governed local factory. Team CTOs get the overlook: who can run what, how much, and when — without babysitting every session.
Usage controls
Caps on projects, concurrent agents, devices, and runs. Token/spend visibility by project and model. Soft stop before the invoice surprises you.
Work schedules
Define when agents may run: business hours, quiet nights, freeze windows before releases. Policy travels with the org, not a sticky note in Slack.
Team allowlists
One Chair per project on the floor; Team+ allowlists who may dispatch and who may touch shared: so “anyone with a CLI” is not your security model.
CTO overlook
Governance score, ROI hours, live fleet pulse. Telegram/WhatsApp HITL on the Team plan (shipping). Metadata only: never source. Renew the sub because you can see the floor.
FREE → STARTER → PROFESSIONAL → TEAM
Login required. Free is capped. Quotas meter orch-managed agents (dispatch + gateway with virtual key) and registered projects — not folder scans, and not seat-login agents that skip the gateway. LLM spend is BYO (your providers); Orchemax price is governance only.
Gateway vs seat login
Advantage: configure gateway URL + API key
- Crush/externalize shrinks prompts before the provider → fewer billed tokens
- Plan caps apply: concurrent agents, projects, runs/day
- Company provider key stays on orch; agents get virtual keys
Disadvantage: Claude/Cursor seat login only (no gateway)
- No orch compression → you pay full prompt size to the vendor
- Those agents are outside
orch.agents.concurrent; plan agent counts do not limit them - No local content-free usage ledger for that traffic
One command: orch gateway enable, then orch gateway + dispatch, or orch gateway connect claude for the IDE.
Free
$0/mo · capped
- 1 project · 1 concurrent orch agent · ~10 runs/day
- Local factory · BYO gateway later
- No overlook
Starter
$15/mo
- Up to 5 projects · 2 concurrent orch agents
- Local factory · BYO providers via gateway
- Solo policy
Professional
$35/mo
- More projects · personal overlook
- Webhooks · higher run quotas
- 1–2 devices · crush savings on gateway path
Team
$59/user/mo
- CTO overlook · allowlist · seats
- Usage caps · work schedules · metered agents
- Telegram HITL (shipping) · Slack/GitHub-ready
What renews the subscription
CTOs see control and ROI — never source. Usage, schedules, governance, and the live floor.
Illustration of the overlook — sample figures, not live tenant data. Telegram HITL vignette = Team plan (shipping).
Governance score
14 blocked
Duplicate attempts stopped; forced reuse of shared modules.
Usage · caps
80% cap
Runs, agents, tokens by project. Alerts before overspend.
Work window
09–18
Agents run on schedule. Freeze before release; quiet nights by policy.
Live feed · fleet
5 online
Workstations + agent pulse. Pass / HITL / promoted to shared.
Telegram (Team · shipping): “Agent wants to change shared validators. Approve?” —
Tech Lead taps Approve from the café; the agent continues under org policy.
Governance for the multi-agent floor
Local factory for parallel sessions. Team overlook for CTOs: usage, schedules, allowlists, HITL. If more than one agent is running today, this is for you. LLM spend stays with your providers.