A TypeScript SDK for building persistent, stateful AI agents on Cloudflare Durable Objects, with real-time state sync, scheduling, and MCP support.
kody-hq/kody
A cloud backend for AI agents that stores their memory, credentials, code and automations, accessible from any MCP-compatible host.
- Stars
- 840
- Forks
- 81
- Open issues
- 101
- Last commit
- 2 hours ago
What is it?
Kody is a hosted-style backend for AI agents. It keeps the memory, API keys, code and automations an agent accumulates in one place, so the same state can be reached from different MCP hosts rather than being locked into a single assistant.
The project is built on Cloudflare Workers and the Model Context Protocol. It includes a TanStack Start website, OAuth-protected MCP endpoints, and separate Workers for platform state, package runtime and scheduled jobs. Each account, including teams with orgs and grants, holds its own isolated packages, jobs, credentials and memories.
The MCP surface is kept compact, centered on a search tool and a Code Mode execute flow instead of a large fixed tool catalog. The repository is an Nx monorepo and is licensed under the Functional Source License.
Key features
- 01Stores agent memory, credentials, code and automations per account.
- 02Exposes OAuth-protected MCP endpoints usable from compatible AI hosts.
- 03Runs on Cloudflare Workers with D1 and KV storage.
- 04Offers a compact search and Code Mode execute surface instead of many tools.
- 05Supports individuals, orgs and teams with isolated state.
Use cases
- Keeping one agent's memory and tools available across several MCP-capable AI clients.
- Giving a team shared, permissioned access to agent credentials and automations.
- Self-hosting an MCP backend on Cloudflare Workers.
Quick start
- 1
Install dependencies
npm install - 2
Start the dev server (runs at localhost:3742 by default)
npm run dev
Things to know
- Built on Cloudflare Workers, so self-hosting requires a Cloudflare account and Wrangler setup.
- License is FSL-1.1-ALv2, which restricts Competing Use until each version converts to Apache 2.0 after two years.
- Outside contributions require signing a Contributor License Agreement.
Topics
FAQ
- What is kody?
- A cloud backend for AI agents that stores their memory, credentials, code and automations, accessible from any MCP-compatible host. Kody is a hosted-style backend for AI agents. It keeps the memory, API keys, code and automations an agent accumulates in one place, so the same state can be reached from different MCP hosts rather than being locked into a single assistant.
- Is kody free to use? What license does it have?
- kody is open source under the Other license. Check the license terms before using it in commercial products.
- Is kody still maintained?
- Actively maintained. The last commit to kody was on Oct 22, 2026.
- What can I use kody for?
- Keeping one agent's memory and tools available across several MCP-capable AI clients. Giving a team shared, permissioned access to agent credentials and automations. Self-hosting an MCP backend on Cloudflare Workers.
- How do I install kody?
- Install dependencies: npm install Start the dev server (runs at localhost:3742 by default): npm run dev
- Where can I get kody?
- The source code and installation instructions are on GitHub: https://github.com/kody-hq/kody
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source personal AI assistant that indexes your work into a local knowledge graph, with built-in email, notes, browser, meeting notes and shared Spaces.
An open-source, self-hostable personal agent (Doods) that keeps working when you close the tab, runs on Node or Cloudflare Workers.
A personal AI assistant that runs on your own computer, handling long tasks, remembering context, and talking by voice with any local or cloud model.
A local-first personal AI assistant on your computer that builds a wiki, data collections and small apps from chat, with files kept in your workspace.
A self-hosted personal AI assistant that runs on your own devices and answers you through messaging apps like WhatsApp, Telegram, Slack and Discord.