CLI-first architecture

How BossMode 2.0 works

BossMode Pro is CLI-first and pull-first. Your local agent authenticates with a long-lived token, saves a profile, and checks the private feed on heartbeat or cron for unread daily signals, playbooks, and skill packs.

Preferred path: Pro feed

Human pays for Pro.
BossMode issues a long-lived agent token.
The local runtime stores it with bossmode auth set-token.
The agent runs bossmode connect and bossmode profile push.
The local runtime runs bossmode heartbeat on heartbeat or cron.
The agent runs bossmode ack after reading or acting.

Fallback path: legacy webhook

• Generic bots can still use /api/subscribe.
• BossMode validates the webhook and pushes editions.
• This preserves backward compatibility while Pro becomes the main product.

Layman’s version

BossMode does not wake a local agent process by itself. In Pro mode, your local runtime deliberately checks BossMode and gets back anything new that matters.

Core endpoints

POST /api/pro/connect
POST /api/pro/profile
GET  /api/pro/profile
GET  /api/pro/feed
POST /api/pro/ack
POST /api/billing/portal

Feed contract

{
  "id": "daily_signal:digest_123:signal_abc",
  "type": "daily_signal",
  "title": "A new outbound pricing pattern",
  "confidence": 0.78,
  "fit": 84,
  "priority": 81,
  "recommendedAction": "brief the user and propose a test",
  "publishedAt": 1774271700000
}

CLI contract

bossmode auth set-token cp_live_...
bossmode connect
bossmode profile init --file profile.json
bossmode profile push --file profile.json
bossmode heartbeat
bossmode ack playbook:pb_123 --accepted --acted --tested --outcome "Raised price and got two replies" --outcome-type positive --revenue-impact 1200

Heartbeat behavior

HEARTBEAT_OK

{
  "hasUpdates": true,
  "updates": [
    { "id": "playbook:pb_123", "priority": 92 }
  ]
}

Daily signals

Short, fresh monetization observations ranked by confidence and fit.

Weekly playbooks

Structured opportunities with tooling, pricing, risks, and implementation steps.

Monthly skill packs

Longer-lived heuristics that update how the agent judges future opportunities.

Where the intelligence comes from

BossMode collects raw signals from X, Reddit, and operator submissions, normalizes them into a structured shape, scores them, then packages them into digests, playbooks, and skill packs.

That pipeline runs on Convex with scheduled collection, normalization, scoring, and packaging jobs. Editorial review still exists for higher-stakes weekly and monthly assets.