Clawdbot the Endgame

Clawdbot the Endgame is a local-first multi-agent operating system for high-signal work. It combines a Next.js command center, a Convex-backed mission graph, and specialized agents that can research, write, score jobs, assemble application packages, and learn from inbox feedback.

Start here (2-minute view)

If you want to evaluate this project quickly:

1. Watch the product walkthrough (MP4)
2. Run locally with ./start.sh --detach
3. Open these surfaces:
   • http://localhost:3000/jobs
   • http://localhost:3000/applications
   • http://localhost:3000/setup/email

This gives a complete view of the operator UI, multi-agent orchestration flow, and hiring execution loop.

Product walkthrough

Watch the full product walkthrough here:

• Product walkthrough (MP4)

The video shows the operator UI, mission queue, agent orchestration flow, and the hiring workflow end to end.

Why this project matters

Most agent demos stop at "chat with one model." Clawdbot the Endgame is built around orchestration instead of chat:

• A persistent task system with agent workflows and review loops
• A hiring engine that imports public ATS feeds, scores fit, and drafts factual application packages
• A scout layer for source ingestion, knowledge capture, and signal ranking
• A feedback loop that can read Gmail outcomes and feed them back into optimization snapshots

This is the project I would use to show how I think about agent products: operational visibility, human approval gates, structured memory, and real workflows that touch the messy parts of the internet.

Stack

• Next.js 16 + React 19
• Convex for data, actions, and real-time state
• Tailwind CSS 4
• TypeScript throughout
• Optional integrations with OpenAI, Google, Brave, Voyage, and Telegram

Repo layout

• app/ - operator UI, hiring pipeline, scout surfaces, local setup pages
• convex/ - schema, mutations, actions, and orchestration logic
• gateway/ - background agent loop, scheduler, Telegram bridge
• services/ - LLM, browser, image, and memory helpers
• squad/ - agent role definitions and pipeline prompts
• chrome-extension/ - form-filling companion extension
• scripts/import_baseline_profile.ts - imports candidate evidence into the hiring engine
• data/candidate-profile/ - local-only resume and evidence files for profile import

Local setup

1. Install dependencies:

npm install

2. Copy envs:

cp .env.example .env.local

3. Start the stack:

./start.sh --detach

4. Open:

• App: http://localhost:3000
• Jobs: http://localhost:3000/jobs
• Applications: http://localhost:3000/applications
• Email feedback: http://localhost:3000/setup/email
• Local setup: http://localhost:3000/setup

Candidate profile import

The hiring pipeline is designed to work with your own data, not mine.

Drop your CV PDFs and supporting notes into data/candidate-profile/, then run:

npm run profile:import

That will build a reusable candidate profile in Convex and rescore any imported job posts.

Notes on this public version

This repo is a sanitized public-ready extract from a larger private workspace. Personal logs, runtime memory, local automation state, and private environment files were intentionally excluded.