AI Agent Army — deploy client-scouting agents from a simple UI

[ceoguy]

What this is

A blueprint and working MVP for an "army" of AI agents that can be deployed from a simple UI to scout for clients across the ChainGPT ecosystem — ChainGPT Pad, ChainGPT AI, and Saleium — with the ability to add new sales motions as data, no redeploy.

It layers a second agent type onto AgenticOS, reusing every primitive the repo already has (ChainGPT Web3 LLM, node-cron, the EJS+Tailwind dashboard, PASSWORD_AUTH middleware, JSON stores).

Mental model

Product (what we sell) ──has──▶ Workflow (who to scout + how to pitch)
Agent = product + workflow + targeting + cron cadence   ← deploy many = the army
Lead  = scored, qualified prospect + personalized outreach draft + pipeline status

Products and workflows are data, not code (data/products.json), seeded with:

• ChainGPT Pad — VC/network scout, IDO clients, launchdrop clients
• ChainGPT AI — Web3 builder scout
• Saleium — public-sale infra scout, vesting/staking portal scout

What ships

• /agents dashboard: deploy → manage (run-now / pause / delete) → triage leads through a new → contacted → qualified → won pipeline.
• Scouting engine (runAgent): builds a brief from the workflow + operator targeting, asks the ChainGPT LLM for scored leads + outreach drafts, persists them.
• Fleet scheduler (agent.job.ts): one cron task per enabled agent, re-applied live on changes.
• Full architecture + roadmap in docs/agent-army-architecture.md.

Extensibility

Add a workflow under a product (or a whole new product) in the registry and it appears in the UI instantly — writing a playbook is a sales exercise (targetPersona, icp, signals, qualificationCriteria, outreachTemplate), not an engineering one.

Notes / next steps

• Discovery is LLM-driven in this MVP (great for qualification + drafting). The runAgent flow is connector-shaped so real signal sources (X search, the existing ChainGPT news webhook, on-chain, CRM sync) slot into the "gather candidates" step without other changes. See doc §6–8.
• Outreach is kept as drafts by default (human-in-the-loop); add opt-in auto-send later.
• LLM endpoint is configurable via the new optional CHAINGPT_LLM_URL env (defaults to the public chat API) — worth confirming against our account's exact endpoint before go-live.

Verification

• tsc --noEmit clean for all new source.
• bun build ./src/index.ts bundles successfully (240 modules).
• Tailwind rebuilt to include the new view's classes.

Opened as a draft for review of the approach and the seeded playbooks.

---

Generated by Claude Code

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 5b1ee80e-1739-42d2-bf42-b81f608d5b22

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/ai-agent-deployment-ui-5HgvN

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}