Davia - Open-Source Interactive Docs from AI Coding Agents

TL;DR

TL;DR: Davia is an open-source CLI tool that lets AI coding agents generate interactive, editable documentation from your codebase, with a Notion-like collaborative workspace for the whole team.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: davia.ai — verified
• Source repository: github.com/davialabs/davia — verified
• License: MIT — verified via GitHub API
• HN launch thread: news.ycombinator.com/item?id=45880716 — verified
• npm package: npmjs.com/package/davia — verified
• Source last checked: 2026-07-01 (README commit verified)

What Is Davia?

Davia is an open-source documentation tool built specifically for AI coding agents. Rather than writing static README files, you give your AI coding agent (Cursor, GitHub Copilot, Windsurf, Claude Code, Augment) the Davia CLI and tools, and it generates interactive documentation for your codebase.

The output is not plain Markdown. Davia produces documentation with interactive visualizations and editable whiteboards — the kind of rich internal docs that help a human understand a codebase at a glance.

The workflow:

1. Install the Davia CLI globally via npm
2. Initialize it with your preferred AI coding agent
3. Ask your agent to write documentation for any part of your project
4. The agent uses Davia’s tools to generate rich, interactive docs
5. View locally via davia open, or push to a shared collaborative workspace

From the README:

Davia is an open-source tool designed for AI coding agents to generate interactive internal documentation for your codebase. When your AI coding agent uses Davia, it writes documentation files locally with interactive visualizations and editable whiteboards that you can edit in a Notion-like platform or locally in your IDE.

Setup Workflow

Prerequisites

• Node.js 18 or higher
• npm or yarn
• An AI coding agent (Cursor, GitHub Copilot, Windsurf, Claude Code, or Augment)

Step 1: Install the Davia CLI

npm i -g davia

Verify the installation:

davia --version

Step 2: Initialize Davia with Your Coding Agent

davia init --agent=[name of your coding agent]

Replace [name of your coding agent] with your agent name, for example:

davia init --agent=cursor
davia init --agent=claude-code
davia init --agent=windsurf

This configures the local Davia environment and registers the necessary tools for your agent to call.

Step 3: Generate Documentation

Ask your AI coding agent to document a file, module, or the entire codebase:

Can you write documentation for the auth module?

The agent invokes Davia tools internally to generate interactive documentation with visualizations.

Step 4: View Your Documentation

Open the generated docs in your browser:

davia open

If the page does not load immediately, refresh the browser.

Step 5: Push to a Collaborative Workspace

Sync your local docs to the cloud workspace for team collaboration:

davia push

This command:

• Opens a browser window for authentication (if not already logged in)
• Creates a new workspace for your project
• Uploads documentation to the cloud
• Opens your workspace in the browser

Note from the README: workspace updates after push are not yet fully supported, but this feature is on the roadmap.

Deeper Analysis

Why Davia Fills a Real Gap

Most AI-generated documentation is flat text. When a new developer joins a project, they get a README with some code snippets — no interactive diagrams, no way to explore component relationships visually.

Davia flips this. The AI agent has structured tools to produce richer output than plain text, and the viewer supports inline visualization. For a complex codebase, that is the difference between a doc that gets skimmed and one that actually gets used.

Editor Compatibility

The generated documentation works in two modes:

• IDE mode: Open the .davia files directly in your editor as structured JSON/YAML, with Davia’s schema
• Workspace mode: Push to the cloud to get the Notion-like collaborative viewer

Technical Stack

Davia is built on Node.js and distributed via npm. The visualization layer supports Excalidraw-style diagrams embedded inline. The workspace is a hosted web application that syncs with local .davia files.

Current Limitations

From the README:

• Pushing to an existing workspace with updates is not yet supported
• The cloud workspace requires authentication via browser OAuth
• No self-hosted option is documented yet

Practical Evaluation Checklist

• Install via npm and run davia --version — should output a version number
• davia init --agent=cursor — should create a .davia config directory
• Verify the .davia/ directory contains expected schema files
• Test davia open — should launch a local server and open the viewer
• Check that the workspace supports Excalidraw-style embedded diagrams
• Confirm MIT license is present in the repository

Security Notes

• davia push requires browser-based OAuth — this is an external authentication flow, not a CLI token
• No self-hosted option means documentation content is stored on Davia’s cloud by default
• For private codebases, evaluate the data handling policy before pushing to the workspace
• The .davia local files are plain structured data — review what your agent writes into them before sharing

FAQ

Q: Which AI coding agents does Davia support? A: As of the current README, Davia officially supports Cursor, GitHub Copilot, Windsurf, Claude Code, and Augment. Support for additional agents may be added.

Q: Can I self-host the Davia workspace? A: No self-hosted option is documented as of this writing. The workspace is a hosted service at davia.ai requiring OAuth login.

Q: What kinds of visualizations does Davia generate? A: Davia supports Excalidraw-style inline diagrams and editable whiteboards embedded within documentation pages.

Q: Is there a way to update an existing workspace? A: Workspace updates after davia push are not yet supported per the README, though this feature is on the roadmap.

Q: What license is Davia published under? A: MIT license, verified via the repository LICENSE file.

Conclusion

Davia solves a specific but increasingly common problem: AI coding agents produce flat, hard-to-navigate documentation that humans struggle to use. By giving the agent structured tools for generating interactive docs — with diagrams, whiteboards, and a collaborative viewer — it raises the bar for what internal documentation can look like.

If your team uses AI coding agents for any significant work, the flat README is already becoming a bottleneck. Davia is worth a weekend afternoon to evaluate properly.

Install it with npm i -g davia, point your agent at a module, and see what it produces.