Built for Agents

Your AI Writes.
You Stay The Author.

Your agent writes directly into your editor. Changes appear as colored highlights. Accept with a keystroke. Reject with another. Every word that ships is a word you approved.

$ npx openwriter install-skill

One command for Claude Code. For other agents, add the MCP config below.

Your Draft. Your Call. Every Time.

Green for inserts. Blue for rewrites. Red for deletes. Hit a to accept, r to reject. Your keyboard. Your judgment. Your draft.

Chapter 1 — Origins.md

The morning air was crisp as she stepped onto the platform.

A thin fog clung to the rails, curling around her ankles like something alive.

She checked her watch — the 7:15 was late again, as it had been every day this week.

It was a Tuesday and Tuesdays were always bad.

The station was emptier than usual. That should have been her first warning.

j k navigate a accept r reject A accept all R reject all

This Is What The Agent Actually Does.

It writes the draft right into your pad. The green highlights are proposals. Everything red is a deletion it's asking for. You stay in the loop on every sentence.

OpenWriter editor showing agent pending changes as green highlights

61 Tools. Zero Copy-Paste.

Your agent drafts, rewrites, organizes, schedules, publishes, and sends — without ever asking you to switch tabs.

Write & Edit

Create documents. Rewrite paragraphs. Insert images. Edit inline text. Version checkpoints so nothing gets lost.

Organize Everything

Spin up workspaces. Nest containers three deep. Move, rename, tag, archive. Your book, your course, your series — structured the way you think.

Publish Anywhere

Post to X, LinkedIn, or GitHub. Schedule with recurring slots. Send newsletters to your subscribers. One agent, every channel.

Import From Anywhere

Pull structured Google Docs. Open raw .md files. Bulk import samples for voice profiling. Your agent ingests, splits chapters, and files it all in the right place.

Generate Images

Agent calls Gemini, drops the image right where you want it. Inline decoration. Cover image. Your call. Your approval.

Manage The Plan

Tasks. Schedules. Autoplugs. Slots. Subscribers. Your agent can run the whole content operation — you just decide what goes live.

View All 61 Tools

Your Agent Writes In Context. Not In A Vacuum.

Workspaces carry the character sheets, the style rules, the series bible. Your agent reads them before it writes a line. That's why it sounds like the book — not like ChatGPT.

✓ Workspace context — characters, settings, rules

✓ Cross-cutting tags — find anything by theme

✓ Agent marks — leave inline feedback for the AI

✓ Version history — nothing you wrote is ever gone

✓ Full-text search across every document

✓ Progressive disclosure — agent fetches what it needs

OpenWriter sidebar showing nested workspaces with containers

 // Select text → right-click → Leave Mark
 
 "tighten this paragraph"
 "is this consistent with Ch 3?"
 "fact-check this claim"
 "give me three alternatives"
 
 // Agent reads the mark,
 // rewrites, resolves it,
 // clears the underline. 

Mark It Up. Let The Agent Handle It.

Select a line. Right-click. Drop a note. The dotted underline tells your agent exactly what you want — no long prompts, no context-switching. The agent addresses every mark on its next turn and clears them as it goes.

✓ Select-based feedback — mark any text range

✓ Notes or silent marks — agent handles either

✓ Dotted underlines visible in editor

✓ Cross-document — see all marks in one view

✓ Auto-resolve after agent addresses them

One Config. Every MCP Agent.

If your agent speaks Model Context Protocol, it works with OpenWriter. One command for Claude Code. Three lines of JSON for everyone else. Real-time WebSocket sync means your agent sees every change the instant you approve it.

✓ One-command install for Claude Code

✓ Standard MCP config for Cursor / Windsurf / VS Code

✓ Real-time WebSocket sync

✓ Multi-session — several agents, different docs

✓ Token-efficient wire — no bloat

View All Tools

 // Add to your MCP config
 {

  "mcpServers": {

    "openwriter": {

      "command": "openwriter",

      "args": ["--no-open"]

    }

  }
 }
 
 // Or for Claude Code:
 // npx openwriter install-skill 

Pick The Agent You Already Trust.

We don't lock you into a model. Bring whichever agent you already use — same tools, same pad, same pending-change review.

Claude Code

One-command skill install. The skill handles the MCP config, launches the server, and opens your editor — all in the terminal.

Cursor

Drop the MCP config into your Cursor settings. Every one of the 61 tools becomes available to your in-editor agent.

Windsurf

Same MCP config. Same toolkit. Your Windsurf agent writes into the same pad, reads the same workspaces, and respects the same pending-change mechanic.

VS Code

Install an MCP extension, plug OpenWriter in, and your editor-bound agent has the whole toolkit — plus live sync to your browser.

Codex, Zed, Copilot

Any agent that speaks MCP. Add the server. You're writing together by lunch.

Multi-Agent Sessions

Two agents. Two documents. One editor. Draft the chapter in one session, schedule tweets in the other. Nothing collides.

The Plugin Ecosystem.

Four extension points. Four bundled plugins. Open for community development. If you want your agent to do something custom, someone's probably already written it.

MCP Tool Registration

Plugins ship their own tools. Your agent picks them up automatically — no config, no restart.

HTTP Routes

Add API endpoints that your agent can hit. Integrate anything — your CRM, your scheduler, your research stack.

Context Menu Actions

Right-click a selection — get plugin-provided options. Agent runs them with the selected text as input.

Sidebar Actions

Document-level plugin commands. Right-click a doc in the sidebar — plugin actions show up alongside the defaults.

Plugin Development Docs

Your AI Writes.You Stay The Author.