MCPWorks Agents
Your AI builds it.
MCPWorks runs it.
Forever.
Persistent, autonomous agents that run 24/7. Describe the automation in natural language. Your AI builds the agent. It runs on a schedule, responds to events, and communicates back to you.
MCPWorks is infrastructure, not AI. You bring your own LLM from 14+ providers — we never host, proxy, or bill for AI tokens.
$ claude "build me an ETH price agent"
✓ Created agent ethwatch
→ ethwatch.agent.mcpworks.io
✓ Created function check-price
→ Python, monitors ETH price feed
✓ Scheduled every 5 minutes
✓ Connected to #alerts on Discord
✓ AI engine: Kimi K2.5 (your API key)
Agent is running. It will message you
on Discord when ETH moves >5%.
3 hours later, on your phone:
ethwatch: ETH dropped 6.8% in the
last hour. Correlated with SEC filing on
staking regulations. Want me to dig
deeper?What is an MCPWorks Agent?
An MCPWorks Agent is a containerized, autonomous entity that runs on your infrastructure or ours. It's not a chatbot. It's not a script. It's a persistent process with its own identity, its own tools, and optionally its own AI brain.
Without MCPWorks
- Rent a VPS, provision infrastructure
- Write cron jobs, manage systemd
- Handle secrets, rotation, encryption
- Build webhook endpoints from scratch
- Wire up Discord/Slack bots manually
- Monitor uptime, restart on crash
- Deploy code changes via SSH
Hours of DevOps for every agent.
With MCPWorks Agents
- "Build me a price monitoring agent"
- Your AI creates the functions
- Sets a cron schedule
- Configures Discord notifications
- Connects its own AI for reasoning
- Agent runs on MCPWorks. Forever.
- You talk to it on your phone.
One conversation. Agent runs 24/7.
The Intelligence Hierarchy
You tell your AI what to build. Your AI writes the high-quality functions and locks them. The agent's cheaper LLM handles day-to-day reasoning. Functions are pure execution.
You
"Build me an agent that monitors competitor pricing and alerts me on Discord."
Your Primary AI (Claude, GPT, Copilot)
Writes high-quality scraping functions, sets up the schedule, configures Discord. Locks the critical functions so the agent can't break them.
MCPWorks Agent (your cheaper LLM)
Runs 24/7 on a schedule. Uses the locked functions for reliable execution. Can create its own ad-hoc functions for new situations. Messages you on Discord when something matters.
Functions (Code Sandbox)
Python or TypeScript executing in nsjail-isolated sandboxes. No AI needed. 70-98% token savings because data stays in the sandbox.
Everything an agent needs
Each agent gets its own container, namespace, and subdomain. Triggers include cron schedules, webhooks, and self-programming heartbeats.
Isolated Containers
Each agent runs in its own Docker container with reserved CPU and RAM. Own namespace, own subdomain for webhooks. Python (59+ packages) or TypeScript (Node.js 20+).
Scheduled Execution
Cron-based scheduling from every 15 seconds to daily. Webhook ingress from Stripe, GitHub, Shopify — anything that sends HTTP. Three orchestration modes: direct, reason-first, and run-then-reason.
Self-Programming Heartbeat
The agent wakes on a configurable interval, loads its goals and soul, and its AI decides what to do. It can rewrite its own instructions for the next wake — a reasoning loop that evolves.
Persistent State & Memory
Encrypted key-value store (AES-256-GCM) per agent. Conversation memory persists across sessions with automatic LLM-driven compaction. Your agent remembers everything.
BYOAI — Your LLM, Not Ours
Connect your own API key per agent from 14+ providers: OpenAI, Anthropic, Google, DeepSeek, Kimi, Groq, and more. Use a cheap model for routine tasks, a powerful one for complex decisions. We never proxy your tokens.
Communication Channels
Agents message you on Discord, Slack, WhatsApp, or email. You reply with instructions from your phone. No IDE needed. Function locking prevents the agent's AI from degrading quality.
Frequently Asked Questions
How is this different from LangChain or CrewAI? +
LangChain and CrewAI give you the SDK — you still need to host, schedule, manage state, handle secrets, and build communication channels yourself. MCPWorks gives you the runtime. Your AI builds the agent through MCP tools, and MCPWorks handles the container, the cron, the webhook ingress, the encrypted state, and the Discord/Slack integration.
What does BYOAI mean? +
Bring Your Own AI. You connect your own LLM API key per agent from 14+ providers. MCPWorks charges for infrastructure — containers, scheduling, state, communication — not for intelligence. We never proxy your AI tokens.
What are heartbeat mode and agent soul? +
Heartbeat mode is proactive autonomy. The agent wakes on a configurable interval and runs its AI reasoning loop — loading its soul (persistent identity), goals, conversation memory, and self-written instructions — then decides what to do. The agent can reprogram its own next wake behavior. Together, heartbeat + soul + memory let an agent pursue goals autonomously with consistent identity and full recall.
Can agents run without AI? +
Yes. Automation mode runs functions on cron schedules and responds to webhooks with no AI engine and $0 token cost. Add an AI engine later when you need reasoning. Start simple, scale up.
Get early access to MCPWorks Agents
Agents are production-ready and available to invited users. Join the waitlist for first access when we expand availability.
No spam. We'll email when agent access opens up.
Free to self-host. Cloud from $179/mo. See pricing →