Skip to content
Back to search

dev-browser-marketplace

MarketplaceWorks with Claude CodeGitHubIncomplete

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/sawyerhood-dev-browser/badge.svg)](/listings/sawyerhood-dev-browser)

HTML

<a href="/listings/sawyerhood-dev-browser"><img src="/api/listings/sawyerhood-dev-browser/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

npm install -g dev-browser
dev-browser install    # only needed if dev-browser cannot find Chrome

The release binary includes Bun and Puppeteer. Node is not required after installation. macOS and glibc Linux are supported; Windows and musl Linux are not yet supported.

Snap-packaged Chromium cannot read the default `~/.dev-browser/v1` directory. On Ubuntu, use `dev-browser install` or point `DEV_BROWSER_CHROME` at another Chrome binary.

If your package manager blocks lifecycle scripts, run `npm rebuild -g dev-browser`, `pnpm approve-builds -g dev-browser`, or `bun pm -g trust dev-browser`. The CLI will also try to download its binary on first use. For a private mirror, set `DEV_BROWSER_DOWNLOAD_BASE`; set `DEV_BROWSER_SKIP_DOWNLOAD=1` if you install the binary yourself.

Quick start

# Launch a headless browser and run a script
dev-browser --headless <<'EOF'
const page = await browser.getPage("main");
await page.goto("https://example.com");
console.log(await page.title());
EOF

# Attach to Chrome started with `dev-browser chrome`
dev-browser chrome
dev-browser --connect <<'EOF'
console.log(await browser.listPages());
EOF

Chrome 136 and newer ignore remote-debugging flags on the default profile. `dev-browser chrome` handles this by using a dedicated profile and checking that Chrome actually started.

Using it with coding agents

Tell the agent to run `dev-browser --help`. The built-in guide covers the current API and the preferred look → act → verify workflow.

Agents that discover local skills can install the bundled skill explicitly:

dev-browser install-skill --codex   # ~/.codex/skills/dev-browser/SKILL.md
dev-browser install-skill --claude  # ~/.claude/skills/dev-browser/SKILL.md
dev-browser install-skill --agents  # ~/.agents/skills/dev-browser/SKILL.md

Run `dev-browser install-skill` without flags to update all three locations.

Idle browser cleanup

Launched browsers close after 30 minutes without a script by default. Change that per command, in the environment, or in `~/.dev-browser/v1/config.json`:

dev-browser --idle-timeout 5m < script.js
DEV_BROWSER_IDLE_TIMEOUT=1h dev-browser -e 'await browser.listPages()'
{
  "idleTimeout": "5m"
}

Durations accept `30s`, `5m`, `1h`, or raw milliseconds. Set the value to `0` to keep a launched browser open until `dev-browser stop`. Attached browsers are never closed by idle cleanup. Profiles, cookies, and login state remain on disk when a launched browser closes.

<details> <summary>Allowing dev-browser in Claude Code without permission prompts</summary>

Add `dev-browser` to the `allow` list in `.claude/settings.json` for one project or `~/.claude/settings.json` for every project:

{
  "permissions": {
    "allow": ["Bash(dev-browser *)"]
  }
}

This allows any matching command without another prompt. Only do this where you trust the scripts being run: `node:vm` gives each script fresh globals, but it is not a security sandbox.

</details>

<details> <summary>Legacy Claude Code plugin installation</summary>

/plugin marketplace add sawyerhood/dev-browser
/plugin install dev-browser@sawyerhood/dev-browser

Restart Claude Code after installation.

</details>

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 29 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "dev-browser-marketplace",
  3. "owner": {
  4. "name": "Sawyer Hood",
  5. "email": "sawyerjhood@gmail.com"
  6. },
  7. "plugins": [
  8. {
  9. "name": "dev-browser",
  10. "source": "./",
  11. "description": "Browser automation skill with persistent page state for developers and AI agents",
  12. "strict": false,
  13. "keywords": [
  14. "browser",
  15. "puppeteer",
  16. "automation",
  17. "testing",
  18. "debugging"
  19. ],
  20. "skills": [
  21. "./skills/dev-browser"
  22. ]
  23. }
  24. ],
  25. "metadata": {
  26. "description": "Marketplace for the dev-browser skill",
  27. "version": "1.0.0"
  28. }
  29. }
README.md · README · 243 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. <p align="center">
  2. <img src="assets/header.png" alt="Dev Browser - browser automation for coding agents" width="100%">
  3. </p>
  4. Brought to you by [Do Browser](https://dobrowser.io).
  5. `dev-browser` lets coding agents control Chrome with short JavaScript scripts. The browser stays open between calls,
  6. so an agent can navigate once, inspect the page, act, and verify the result without starting over each time.
  7. **Key features:**
  8. - **Persistent pages.** Named tabs carry state across scripts.
  9. - **Compact snapshots.** Accessibility trees give agents readable output and stable element refs.
  10. - **Real Puppeteer.** Scripts use the Puppeteer Page API plus a small set of agent-focused helpers.
  11. - **Launch or attach.** Start an isolated Chrome profile or connect to a browser you already have open.
  12. ## Demo
  13. https://github.com/user-attachments/assets/c6cf7fb9-b1dc-46ed-93b9-6e7240990c53
  14. ## CLI installation
  15. ```bash
  16. npm install -g dev-browser
  17. dev-browser install # only needed if dev-browser cannot find Chrome
  18. ```
  19. The release binary includes Bun and Puppeteer. Node is not required after installation. macOS and glibc Linux are
  20. supported; Windows and musl Linux are not yet supported.
  21. Snap-packaged Chromium cannot read the default `~/.dev-browser/v1` directory. On Ubuntu, use `dev-browser install`
  22. or point `DEV_BROWSER_CHROME` at another Chrome binary.
  23. If your package manager blocks lifecycle scripts, run `npm rebuild -g dev-browser`, `pnpm approve-builds -g
  24. dev-browser`, or `bun pm -g trust dev-browser`. The CLI will also try to download its binary on first use. For a
  25. private mirror, set `DEV_BROWSER_DOWNLOAD_BASE`; set `DEV_BROWSER_SKIP_DOWNLOAD=1` if you install the binary yourself.
  26. ### Quick start
  27. ```bash
  28. # Launch a headless browser and run a script
  29. dev-browser --headless <<'EOF'
  30. const page = await browser.getPage("main");
  31. await page.goto("https://example.com");
  32. console.log(await page.title());
  33. EOF
  34. # Attach to Chrome started with `dev-browser chrome`
  35. dev-browser chrome
  36. dev-browser --connect <<'EOF'
  37. console.log(await browser.listPages());
  38. EOF
  39. ```
  40. Chrome 136 and newer ignore remote-debugging flags on the default profile. `dev-browser chrome` handles this by using
  41. a dedicated profile and checking that Chrome actually started.
  42. ### Using it with coding agents
  43. Tell the agent to run `dev-browser --help`. The built-in guide covers the current API and the preferred
  44. look → act → verify workflow.
  45. Agents that discover local skills can install the bundled skill explicitly:
  46. ```bash
  47. dev-browser install-skill --codex # ~/.codex/skills/dev-browser/SKILL.md
  48. dev-browser install-skill --claude # ~/.claude/skills/dev-browser/SKILL.md
  49. dev-browser install-skill --agents # ~/.agents/skills/dev-browser/SKILL.md
  50. ```
  51. Run `dev-browser install-skill` without flags to update all three locations.
  52. ### Idle browser cleanup
  53. Launched browsers close after 30 minutes without a script by default. Change that per command, in the environment, or
  54. in `~/.dev-browser/v1/config.json`:
  55. ```bash
  56. dev-browser --idle-timeout 5m < script.js
  57. DEV_BROWSER_IDLE_TIMEOUT=1h dev-browser -e 'await browser.listPages()'
  58. ```
  59. ```json
  60. {
  61. "idleTimeout": "5m"
  62. }
  63. ```
  64. Durations accept `30s`, `5m`, `1h`, or raw milliseconds. Set the value to `0` to keep a launched browser open until
  65. `dev-browser stop`. Attached browsers are never closed by idle cleanup. Profiles, cookies, and login state remain on
  66. disk when a launched browser closes.
  67. <details>
  68. <summary>Allowing dev-browser in Claude Code without permission prompts</summary>
  69. Add `dev-browser` to the `allow` list in `.claude/settings.json` for one project or `~/.claude/settings.json` for every
  70. project:
  71. ```json
  72. {
  73. "permissions": {
  74. "allow": ["Bash(dev-browser *)"]
  75. }
  76. }
  77. ```
  78. This allows any matching command without another prompt. Only do this where you trust the scripts being run:
  79. `node:vm` gives each script fresh globals, but it is not a security sandbox.
  80. </details>
  81. <details>
  82. <summary>Legacy Claude Code plugin installation</summary>
  83. ```text
  84. /plugin marketplace add sawyerhood/dev-browser
  85. /plugin install dev-browser@sawyerhood/dev-browser
  86. ```
  87. Restart Claude Code after installation.
  88. </details>
  89. ## Script API
  90. Scripts get these globals:
  91. ```javascript
  92. // Browser control
  93. browser.getPage(nameOrId) // Get/create a named page, or attach by target ID
  94. browser.newPage() // Create an anonymous page; close it yourself
  95. browser.listPages() // [{ id, url, title, name }]
  96. browser.closePage(name) // Close a named page
  97. // File I/O, restricted to ~/.dev-browser/v1/tmp
  98. saveFile(name, data)
  99. readFile(name)
  100. // Output
  101. console.log()
  102. console.warn()
  103. console.error()
  104. ```
  105. Top-level `await` works, and the last expression becomes the command result. Pages are real [Puppeteer Page
  106. objects](https://pptr.dev/api/puppeteer.page) with a few additions:
  107. ```javascript
  108. await page.snapshot({ interactive: true }) // Accessibility tree with refs such as e12
  109. await page.click("ref/e12") // Refs work in Puppeteer selector methods
  110. await page.ref("e12") // ElementHandle for a ref
  111. await page.shot() // JPEG path and CSS-pixel dimensions
  112. await page.waitForLoad() // Wait for navigation, requests, and DOM activity to settle
  113. await page.fill("#email", "me@example.com")
  114. ```
  115. Each command runs in a fresh `node:vm` context inside the daemon. This keeps script globals separate, but it is not a
  116. security boundary. When a command finishes or times out, its page, locator, browser-context, and registry operations
  117. are closed so detached work cannot interfere with the next command.
  118. Scripts may run concurrently. Browser and page creation are serialized, as are input operations on different tabs
  119. through a bring-to-front lock. Two scripts using the same named page can still interleave.
  120. See [`dev-browser --help`](docs/help.md) for the full API, error behavior, configuration, JSON output, MCP tools, and
  121. examples.
  122. ## Connecting to an existing browser
  123. `--connect` accepts auto-discovery, a port, an HTTP URL, a WebSocket URL, or a raw CDP Unix socket:
  124. ```bash
  125. dev-browser chrome --profile work
  126. dev-browser --connect -e 'await browser.listPages()'
  127. dev-browser --connect 9222 -e 'await browser.listPages()'
  128. dev-browser --connect 'wss://provider.example?token=…' -e 'await browser.listPages()'
  129. ```
  130. Attached browsers belong to the user. dev-browser extends only the tabs a script asks for and never closes the browser
  131. because of an idle timeout. Credentials are redacted from logs and status output, while differently authenticated
  132. endpoints remain separate sessions.
  133. ## MCP
  134. ```bash
  135. claude mcp add dev-browser -- dev-browser mcp --headless
  136. ```
  137. The server exposes `dev_browser_run`, `dev_browser_pages`, `dev_browser_browsers`, `dev_browser_stop`, and
  138. `dev_browser_help` over the same warm daemon as the CLI.
  139. ## Upgrading to 1.0
  140. Version 1.0 replaces the Playwright/QuickJS implementation from dev-browser 0.2 with the Puppeteer/Bun implementation
  141. developed as doobie. Its state lives under `~/.dev-browser/v1`, separate from both older installations.
  142. To copy durable doobie state:
  143. ```bash
  144. doobie stop
  145. dev-browser migrate-from-doobie
  146. ```
  147. `migrate-from-doobie` leaves `~/.doobie` untouched. Scripts written for dev-browser 0.2 need to replace helpers such as
  148. `snapshotForAI()` and `getByRef()` with `snapshot()` and `ref/eN`. Computer-use helpers under `page.cua` and
  149. `page.domCua` are not part of 1.0.
  150. ## Benchmarks
  151. Measured on Linux with a headless browser and warm daemon, medians of 9 runs:
  152. | Scenario | Time |
  153. | --- | ---: |
  154. | Empty script | ~13 ms |
  155. | `getPage("x")` + `page.title()` | ~14 ms |
  156. | `page.snapshot()` on a SERP-like page | ~18 ms |
  157. | `page.shot()` | ~36 ms |
  158. | Cold daemon and Chrome launch | ~280 ms |
  159. Run them with `bun run build && bun run bench/run.ts --runs 9`. The older end-to-end comparison is in
  160. [dev-browser-eval](https://github.com/SawyerHood/dev-browser-eval).
  161. ## Development
  162. ```bash
  163. bun install
  164. bun run dev -- -e '1+1'
  165. bun test
  166. bun run build
  167. ```
  168. Design notes live in [docs/design-decisions.md](docs/design-decisions.md).
  169. ## Releasing
  170. See [RELEASING.md](RELEASING.md) for release-candidate testing, npm trusted publishing, and post-release verification.
  171. ## License
  172. MIT
  173. ## Author
  174. [Sawyer Hood](https://github.com/sawyerhood)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.