Your coding agent can write books now.
Authr is an MCP server with a manuscript attached. Connect Claude Code, Claude Desktop, Cursor or your own agent and it gets 100 tools: read the book, draft scenes, file inline suggestions, comment, update the codex and timeline, run brainstorms. You review in a calm editor built for prose.
- Add Authr
- Sign in
- Ask for work
Connectors
Allow Claude to reference other apps and services for more context.
Connectors
Use connectors to let ChatGPT search and work with your apps.
claude mcp add --transport http authr https://app.authr.shop/mcp --header "Authorization: Bearer bk_…"Make a key in the app at Settings → API keys, then replace bk_…. Full walkthrough: Write a book with Claude Code.
One command to connect
Add the server to Claude Code with your API key. Every route is a tool with a summary and typed parameters, so your agent finds its own way around.
Suggestions, not overwrites
Give an API key a role with propose-only permissions and every change it makes to existing prose becomes an inline suggestion you review. Drafting empty scenes applies directly.
REST for everything else
The same 100 actions at /v1, documented for agents at /llms.txt. Keys can be locked to a single role.
Webhooks for your own bots
Point a role at your endpoint and it's called on every @mention or delegated task with a full briefing. Answer in the HTTP response and the reply lands in the thread. Deliveries retry on failure.
An events feed
Poll /v1/events to see changes as they happen: new comments, accepted suggestions, finished runs.
Free for external agents
Agents you connect run on your own AI, so they're unlimited on every plan. In-app agent tokens are only for agents that run inside the app.
Questions
Which MCP transport does it use?
Streamable HTTP with stateless JSON responses, authenticated with a Bearer API key.
Can I run my own model?
Yes, as an external agent: connect it over MCP, call the REST API, or receive webhooks. It runs wherever you like.