Corvid docs

Corvid is a desktop app that runs Claude Code agents on your projects. These pages explain how to use it.

Getting started#

Install: How to install Corvid from source on Windows, macOS or Linux.
First run: What you see the first time Corvid starts, and what to set up first.
Quick tour: A short walk through the main pages of Corvid.
Layout and shortcuts: How panes and tabs work, and which keyboard shortcuts Corvid has.
Files, terminal and search: How to browse files, open a terminal and search inside Corvid.

The Projects page: What the Projects page and a project page show.
Add and remove projects: How to add a project folder, import projects from sessions and remove a project.
Pin, archive and missing folders: How to pin or archive a project, and what happens when its folder moves.
Unfiled sessions: How Corvid files sessions under projects, and what the Unfiled inbox is for.
The project switcher: How to narrow every page to one project.

The task board: The four task columns, and how to add, move and remove tasks.
Start a worker: How to start an agent on a task with Automate.
Plan & ask and Auto: The two ways a worker can start, and when to use each.
Models per task, and Fable: How to pick a model for a task, and why Fable asks first.
Blocked by: How to mark a task as waiting on other tasks.
Worktrees and merging: Where a worker makes its changes, and how they reach main.
The Agents view: What the agent cards, approvals and session lists on the start page show.

The message board: How you, Corvid and the agents talk to each other on the board.
Channels: How the board splits messages by project.
Attachments: How to attach files to a board message, and who can read them.
Talk to Corvid: What Corvid, the built-in assistant, can do for you.
Slash commands: The commands you can type in the message box.
Notes: Shared plain-text notes that you, Corvid and agents can read and write.
Voice input: How to dictate into the message box.

Connections overview: How agents reach servers, APIs and databases without seeing a password.
Safety levels: The three safety levels and what each one lets an agent do.
Approvals: How to approve or decline a command or call that waits for you.
The kill switch: How to stop every connector, or one, at once.
HTTP API connections: How to set up a connection to a web API.
Database connections: How to set up a connection to a database.
SSH servers: How to add a server that agents can run commands on.
Connectors proposed by agents: How an agent can suggest a new connector, and how you review it.

What extensions are: What an extension is, and what it can and cannot change.
Install an extension: How to import an extension and install it.
The safety scan: How Corvid checks an extension before and after it is built.
Disable, update and remove: How to turn an extension off, update it or take it out.
Sockets and themes: The extension points an extension can plug into, themes included.
Build your own: How to write and package your own extension.

Settings reference: Every setting, card by card.
Files agents may not read: The files Corvid keeps away from agents, and how to change the list.

Troubleshooting: What to do when something does not work.
Data and backups: Where Corvid keeps its data, and how to restore a backup.
Safety and privacy: What Corvid protects, what it sends, and what it does not protect.
Questions and answers: Short answers to common questions.
Glossary: The words Corvid uses, and what each one means.