Monkeey
Team led by an Associate Founder (Safoan Touil) — hands-on Web3/DeFi engineer (TypeScript, JavaScript, Solidity), YC Startup School alum, Monadex Labs lead.
Project Description
SINCE I was not able to paste the loom link on project video I post it here : https://www.loom.com/share/9aa3ab4a1f9947dc80ae815a4f2765ed
Monkee: Autonomous AI DevOps for Solo Founders
Monkee eliminates DevOps headaches for vibe coders. Connect any
GitHub repository, and our AI agents automatically analyze your codebase, detect
frameworks, generate optimized Dockerfiles, and deploy to Google Cloud Run.
When builds fail from missing dependencies or config errors, Monkee’s self-healing
system autonomously diagnoses issues, modifies your code, and retries deployment, no
manual debugging needed.
Post-deployment, monitoring agents check your app’s health every 30 seconds. When
crashes occur, Monkee fetches error logs, applies fixes, and redeploys automatically
while you sleep.
Real-time dashboard shows live monitoring status, response time charts, and auto-fix
activity. It’s your AI DevOps engineer that never takes a break.
partner tech used
Mistral as llm layer
Google cloud
Qdrant
Typescript
Langgraph for agents orchestrations
express
Nextjs
Mistral
// agents logic
Autonomous Agent Logic:
Self-Healing Workflow:
- Deploy → triggerDeploymentWorkflow(config)
- IF deployment fails:
a. detectAndClassifyError() → { errorType, confidence }
b. IF not fixable OR confidence < 60% → escalate to human
c. diagnoseError() → { rootCause, fixStrategy, confidence }
d. IF diagnosis confidence < 70% → escalate
e. applyFixes() → modify files (package.json, Dockerfile, etc.)
f. Retry deployment (attempt 2/3)
- IF success → startMonitoring()
- Monitoring loop (every 30s):
a. checkHealth() → { healthy, responseTime, statusCode }
b. IF 3 consecutive failures → detectCrash()
c. fetchCloudRunLogs() → error stack traces
d. Run same fix workflow (detect → diagnose → fix → redeploy)
Failure Modes & Safeguards:
- Low Confidence Fixes: If diagnosis confidence < 70%, agent stops and alerts user
instead of applying risky fixes
- Max Retry Limit: Only 3 auto-fix attempts to prevent infinite loops
- Fixable Error Whitelist: Only auto-fixes known error types (dependencies,
scripts, env vars, ports) - unknown errors escalate
- Graceful Degradation: If Qdrant/vector DB fails, system continues without
knowledge base
- Logging: All fix attempts stored in monitoring events for debugging
Monkeey is super scalable and is inspired from lovable growth strategy, a product that seems basic but super scalable where adoption can syrocket, It’s the only deployment platform where AI doesn’t just assist - it fully operates. You’re not clicking through suggestions or copy-pasting fixes. Monkee autonomously detects failures, reasons about solutions, modifies your code, and redeploys. It’s DevOps on autopilot for the solo founder who wants to build products, not babysit infrastructure. 🚀🐒
Prior Work
0 prior work