# autoplans.dev > Autoplans turns a described project into a structured, editable backlog of tasks and subtasks, then exposes that backlog to coding agents over the Model Context Protocol. It runs as a hosted MCP server, a VS Code extension, an Obsidian plugin, a desktop app and a CLI, so the plan lives in one place and every agent reads the same one. - Hosted MCP endpoint: `https://autoplans.dev/api/v1/mcp` (Streamable HTTP, `Authorization: Bearer apk_live_...`). - The MCP server exposes around forty tools covering projects, tasks, subtasks, comments, assignment, branding, business plans, code-agent helpers and telemetry. - For clients that only speak stdio, the `autoplans-mcp` npm package bridges to the same endpoint. - First-party clients: a VS Code extension, an Obsidian plugin, a Claude Code plugin, an OpenCode plugin, a desktop app and a CLI. All of them are clients of the same account. - Autoplans is bring-your-own-key. Adding your own Anthropic, OpenAI, Google or OpenRouter key gives unlimited AI on every plan, including Free; the provider bills you directly with no markup. - Paid plans start at $19/month and buy orchestration — autonomous task execution, the hosted MCP server, durable workflows, analytics — not tokens. - A plan syncs to a `.autoplans` folder in the repository, so the backlog can live in git next to the code it describes. ## Start here - [Introduction](https://autoplans.dev/docs.md): What Autoplans is and how its pieces fit together. - [Quickstart](https://autoplans.dev/docs/quickstart.md): From an empty folder to a working backlog in ten minutes. - [Accounts and API keys](https://autoplans.dev/docs/accounts.md): Signing in, how each app authenticates, and where credentials are stored. ## Concepts - [Projects](https://autoplans.dev/docs/projects.md): The unit everything else hangs from, and how a folder is linked to one. - [Tasks](https://autoplans.dev/docs/tasks.md): Status, priority, subtasks, dependencies, comments, and linked commits. - [The task map](https://autoplans.dev/docs/task-map.md): The project drawn as a graph, on every surface. - [Agents](https://autoplans.dev/docs/agents.md): The five Autoplans agents and the method they follow. - [Sessions](https://autoplans.dev/docs/sessions.md): How coding sessions work and how to continue one from another tool. ## Apps - [Desktop app](https://autoplans.dev/docs/desktop.md): The coding agent as a desktop application. - [CLI](https://autoplans.dev/docs/cli.md): The same agent in a terminal. - [VS Code extension](https://autoplans.dev/docs/vscode.md): Projects, tasks, and the map inside your editor. - [Obsidian plugin](https://autoplans.dev/docs/obsidian.md): Your backlog in your vault, synced both ways. - [Chrome extension](https://autoplans.dev/docs/chrome.md): Record narrated product demos, with sensitive content hidden as it happens. - [Web dashboard](https://autoplans.dev/docs/dashboard.md): Planning, branding, business plans, and the code agent in the browser. ## Integrations - [MCP server](https://autoplans.dev/docs/mcp.md): Connect any MCP client to your Autoplans account. - [Claude Code](https://autoplans.dev/docs/claude-code.md): The Autoplans plugin and marketplace for Claude Code. - [OpenCode](https://autoplans.dev/docs/opencode.md): The opencode-autoplans plugin. - [GitHub](https://autoplans.dev/docs/github.md): Linking commits and pull requests to tasks. ## Reference - [MCP tools](https://autoplans.dev/docs/tools.md): Every tool the server exposes, with its arguments. - [Models and the router](https://autoplans.dev/docs/models.md): Providers, subscriptions, and how a model is chosen. - [Configuration](https://autoplans.dev/docs/configuration.md): Files, environment variables, and what overrides what. - [Troubleshooting](https://autoplans.dev/docs/troubleshooting.md): What to check when something does not work. ## Guides - [The Model Context Protocol, explained for people who have to wire it up](https://autoplans.dev/guides/model-context-protocol.md): What an MCP server is, how JSON-RPC tool calls work, the difference between stdio and Streamable HTTP transport, and what a client actually sends the model. - [MCP servers worth connecting to a coding agent](https://autoplans.dev/guides/mcp-servers-for-developers.md): The servers that do real work in a Claude Code or Cursor setup — repositories, databases, browsers, files, and the project backlog — with what each one exposes. - [Why a coding agent needs a backlog, not a prompt](https://autoplans.dev/guides/ai-coding-agent-backlog.md): Long-horizon tasks, multi-agent sessions and sprint planning all fail the same way when the plan lives in a context window. What to give the agent instead. - [AGENTS.md: what to put in it, how to structure it, and when to use a skill instead](https://autoplans.dev/guides/agents-md.md): A working template for the instruction file OpenCode, Codex and other agents read first, the sections that earn their place, and the ones that rot. - [.cursorrules and .cursor/rules: examples that hold up](https://autoplans.dev/guides/cursorrules.md): How Cursor reads its rules files, a template for a Next.js project, what to generate versus write by hand, and how to keep the rules from going stale. - [GitHub Copilot custom instructions: configuring copilot-instructions.md](https://autoplans.dev/guides/copilot-instructions.md): Where the file goes, what Copilot does with it, an example for a TypeScript repository, prompt files, and how to generate one from the project itself. - [CLAUDE.md: templates and the project context Claude Code actually uses](https://autoplans.dev/guides/claude-md.md): How Claude Code loads CLAUDE.md, a template for a Next.js project, what belongs in it, and how to point it at the plan instead of copying the plan into it. - [Preventing AI context rot: why agents follow outdated instructions](https://autoplans.dev/guides/context-rot.md): Instruction files describe a moment and the project moves on. How to tell what rots, how to keep the durable part small, and where to keep the part that changes. ## Connecting an MCP client - [MCP server](https://autoplans.dev/mcp-server): the endpoint, the tools, and configuration for every client. - [Claude Code](https://autoplans.dev/mcp-server/claude-code): Add the Autoplans MCP server to Claude Code so the agent reads its tasks, checks dependencies and updates status from the terminal. - [Cursor](https://autoplans.dev/mcp-server/cursor): Add the Autoplans MCP server to Cursor with one mcp.json block, so the agent can read and update the project backlog from the editor. - [VS Code and GitHub Copilot](https://autoplans.dev/mcp-server/vscode-copilot): Connect GitHub Copilot Chat in VS Code to the Autoplans MCP server, with the API key prompted for rather than committed. - [Claude Desktop](https://autoplans.dev/mcp-server/claude-desktop): Add Autoplans to claude_desktop_config.json through the autoplans-mcp stdio bridge, so a Claude Desktop conversation can read and update the backlog. - [Windsurf](https://autoplans.dev/mcp-server/windsurf): Add the Autoplans MCP server to Windsurf Cascade through mcp_config.json, using the serverUrl field Cascade expects for a remote server. - [Cline](https://autoplans.dev/mcp-server/cline): Add the Autoplans MCP server to Cline as a streamableHttp server, edited from the MCP Servers panel rather than by path. - [OpenCode](https://autoplans.dev/mcp-server/opencode): Connect OpenCode to Autoplans, either through the opencode-autoplans plugin or as a remote MCP server in opencode.json. ## Optional - [Everything above as one file](https://autoplans.dev/llms-full.txt): the complete documentation, concatenated. - [VS Code extension](https://autoplans.dev/vscode): AI project management and a task backlog inside VS Code. - [Obsidian plugin](https://autoplans.dev/obsidian): Markdown task management, synced both ways, and the vault as an MCP server. - [Apify Autoplans](https://autoplans.dev/apify-autoplans): the AI web scraping assistant for VS Code. - [Pricing](https://autoplans.dev/pricing): plans, what bring-your-own-key covers, and managed credits. - [Comparisons](https://autoplans.dev/compare): how Autoplans differs from issue trackers built for people.